Chapter 2: Characterization of Communication

Signals and Systems

Goals of the Chapter
 Signals can be categorized in a number of ways
 Will cover characterization of signals and systems
encountered in digital communication systems
 Focus points
 Characterization of bandpass signals and systems
 Vector space representation of signals
 Representation of digitally modulated signals

2
Overview
 Signals and systems
 Stationary stochastic process
 Signal space
 Spectral Characteristics of Digitally Modulated Signals

Digital Comm. – 3
Communication Signals & Systems Characterization
 Binary bits from the source encoder mapped into signal
waveforms, mostly after channel encoding
 In channel encoding redundancy is added in a controlled manner
for error correction at the receiver
 Example: Binary modulation: 0 → s1(t) and 1→ s2(t)
 b bits at a time mapped using M= 2b waveforms
si(t), i = 0,1,2,…M-1
 I.e., one waveform for each of the 2b possible bit sequences
 M-ary modulation for M > 2
 Question: What should be the characteristics of these
waveforms and how do we describe and use them?
 Various forms of digitally modulated signals will be introduced
along with their spectral and other characteristics

Digital Comm. – 4
Bandpass Signals and Systems
 Channels have limited bandwidth centered about the
carrier (DSB) or adjacent to the carrier (SSB)
 Narrowband Definition: BW<< fc, carrier frequency
 Reduce all bandpass signals and channels to equivalent
lowpass signals and channels
 Without any loss of generality, it makes the analysis
independent of the carrier frequency
 We consider that s(t) has a narrowband frequency content
in the vicinity of the frequency, fc

Digital Comm. – 5
Bandpass Signals and Systems
 Construct an analytic signal that contains only the positive
frequencies

S  (t )  S (t )  j S (t ) OR S  ( f )  2U ( f ) S ( f )

 1 s ( )
Where S (t )  
  t  
d , Hilbert transform of s ( )

 The analytic signal S+(t) is a bandpass signal

 Equivalent lowpass representation can be obtained by
frequency translation
  S (t )  j S (t ) e  j 2f c t
 
 j 2f c t
S l (t )  S  (t ) e OR
 

S (t )  j S (t )  S l (t )e j 2f c t

6
Bandpass Signals and Systems …
 Since Sl(t) is in general complex, it may be expressed as
Sl (t)  x(t) jy(t) then

S(t) j S (t)   x(t) jy(t)cos 2 πf c t  jsin 2 πf c t 
  x(t) cos 2 πf c t  y(t) sin 2 πf c t  
j  x(t) sin 2 πf c t  y(t) cos 2 πf c t 

 Equating the real and imaginary parts

S (t )  x(t ) cos 2f c t  y (t ) sin 2f c t

(1)
S (t )  x(t ) sin 2f c t  y (t ) cos 2f c t

 S(t) is the desired form of the bandpass signal

 x(t) and y(t) are amplitude modulations impressed on carriers
cos 2πfct and sin2πfct, which are in phase quadrature

7
Bandpass Signals and Systems …
 Alternatively,

  
S(t)  Re x(t) jy(t e j 2 πf ct  Re Sl (t) e j 2 πf ct  (2)

 A third possible form of representation of S(t) can be

Sl (t)  a(t) e jθ (t) Where

1 y(t) (3)
a(t)  (x (t)  y (t))
2 2
and θ(t)  tan
x(t)
 
S(t)  Re Sl (t) e j 2 πf ct  a(t) cos 2 πf ct  θ(t)

 a(t) and θ(t) are the envelope and the phase angle of S(t),
respectively

8
Bandpass Signals and Systems …
 The signal can also be expressed in frequency domain
through its Fourier transform
 
S ( f )   S (t ) e j 2ft dt   Re S 
l (t ) e j 2f ct
e j 2ft
dt
 

 And using Re ( Z )  1 Z  Z * 
2


1

 We get S ( f )   Sl (t ) e j 2f ct  Sl* (t ) e  j 2f ct e  j 2ft dt
2 


1
2

S l ( f  f c )  S l* ( f  f c ) ;
 Where Sl(f) is the transform of Sl(t)
9
Bandpass Signals and Systems …
 The energy in the signal S(t) is given by
 
ε   S 2 (t) dt   ReS (t)e  dt
i 2 πf c t 2
l
 
 
1 1
  Sl (t) dt   Sl (t) 2 cos(4 πft  2 θ(t) ) dt
2

2  2 

 Since in the second term │Sl(t)│ varies very slowly its

contribution may be neglected and the energy of the signal
given by the first element of the sum only

1
ε   Sl (t ) 2 dt
2 

10
Bandpass Signals and Systems …

11
Linear Bandpass System
 Linear bandpass systems are characterized by the
impulse response h(t) or by the frequency response H(f),
which is the Fourier transform of h(t)
 Note that for real h(t), H*(-f)= H(f)
 Define:
H ( f ) f 0
H l ( f  fc )  
0 f 0

 Then:
0 f 0
H ( f  f c )  
*

f 0
l
H ( f )
 H ( f )  H l ( f  f c )  H l* ( f  f c )

12
Linear Bandpass System
 Or in time domain using the inverse transform

h(t)  hl (t)e j 2 πf ct  hl* (t)e  j 2 πf ct


 2 Re hl (t)e j 2 πf c t

 Where hl(t) and Hl(f) are Fourier transform pairs and are in
general complex valued functions that characterize the
equivalent lowpass system

13
Response of Bandpass System to Bandpass Signal
 Assume an input signal S(t) is a narrowband bandpass
(BP) signal and the system is also narrowband BP

S(t) r(t)
h(t) or hl(t)
Sl (t) rl (t)

r(t)  Re (rl (t)e j 2 πf c t )




 Where r(t)  S(τ ) h(t  τ) dt or

R(f) S(f)H(f)

1
R( f )  [ Sl ( f  f c )  S*l ( f  f c )] [ H l ( f  f c )  H*l ( f  f c )]
2
14
Response of Bandpass System to Bandpass Signal
 For narrowband signal and narrowband impulse response

Sl ( f  f c )  0 and H l ( f  f )  0 for f  0
Thus Sl ( f  f c ) H l* ( f  f c )  Sl* ( f  f ) H l ( f  f )  0

1
R( f )  [ sl ( f  f c ) H l ( f  f c )  Sl* ( f  f c ) H l* ( f  f c )]
2
1
 [ Rl ( f  f c )  Rl* ( f  f c )]
2
 Where
Rl ( f )  Sl ( f ) H l ( f )
 is output spectrum of the LPF system excited by LP signal
15
Overview
 Signals and systems
 Stationary stochastic process
 Signal space

16
Bandpass Stationary Stochastic Processes
 Suppose n(t) is a sample function of a wide sense
stationary (WSS) stochastic process with zero mean and
power spectral density Φnn(f)
 Φnn(f) is assumed zero outside an interval Δf centered around  fc
 n(t) is narrowband process if Δf << fc
 n(t) may be represented by any of the following three forms
n(t)  a(t) cos 2 πf ct  θ(t)
 x(t) cos 2 πf ct  y(t) sin 2 πf ct

 Re z(t) e  j 2 πf ct 
 Where
z(t)  x(t)  jy(t)
y (t )
 (t )  tan 1 and a(t)  x 2 (t )  y 2 (t )
x(t )

17
Bandpass Stationary Stochastic Processes …
 Since n(t) is zero-mean, x(t) and y(t) are also zero mean
 Furthermore, from the stationarity of n(t) follows
xx ( )   yy ( ); xy ( )   yx ( ) and
nn ( )  xx ( ) cos 2f c   yx ( ) sin 2f c

 Which is identical in form with the expression for n(t)

 The autocorrelation of the equivalent lowpass process
z (t )  x(t )  jy(t ) is
1
 
zz ( )  E z  (t ) z (t   )  xx ( )  j yx ( )
2
 Which is the autocorrelation of the complex evelope
 Finally, nn ( )  Re  zz ( )e j 2f  
c

18
Bandpass Stationary Stochastic Processes …
 Thus Φnn() of the bandpass stochastic process is
uniquely determined from the autocorrelation function
ΦZZ() of the equivalent lowpass process z(t) and carrier
frequency fc
 Note that


 Re  e  j 2 f 
j 2f c
nn ( f )  zz ( )e d



1
zz ( f  f c )  zz ( f  f c )
2
 Where ΦZZ(f) is the power spectrum of the lowpass
process z(t)
 Since ΦZZ() = ΦZZ*(-), it follows that ΦZZ(f) is real valued
function of frequency
19
Bandpass Stationary Stochastic Processes …
  xy ( ) is an odd function of  and  xy (0)  0 and hence x(t)
and y(t) are uncorrelated for  = 0
 If n(t) is Gaussian x(t) and y(t) are jointly Gaussian and for
 = 0 are independent

20
Representation of White Noise
 White noise is wideband and cannot be represented in
terms of quadrature components
 If the noise is assumed to have passed through an ideal
bandpass filter, the output can be represented by
quadrature components
Φnn(f)

1/2N0

B B
White BP white
noise noise
-fc fc

21
Representation of White Noise …
 The equivalent lowpass noise z(t) has a power spectral
density:
 1
 N0 f  B
 zz ( f )   2
1
0 f  B
 2
 And autocorrelation function is:
sin B
 zz ( )  N 0 which  N 0 ( ) as B  
B
 yx ( )  0 for all  and  zz ( )   xx ( )   yy ( )

 i.e the quadrature components x(t) and y(t) are

uncorrelated for all time shifts  and the autocorrelation of
z(t), x(t) and y(t) are all equal

22
Overview
 Signals and systems
 Stationary stochastic process
 Signal space

23
Signal Space Concepts
 Analogous to space vectors, we represent a family of
signals such as X   x0 (t ), x1 (t ),..........xM 1 by a signal
space over a given time interval
 A signal space is defined by its orthonormal basis
f 0 (t ), f1 (t ),.................f N 1 
 The inner product of two signals is defined as
b
x1(t) , x2 (t)   x1(t) x*2 (t) dt
a

 The norm of a signal is defined as

1
 
b 2

  x(t) dt 
x(t)  
2

a 
24
Signal Space Concepts …
 A set of signals are orthonormal iff they are orthogonal and
their norms are each unity
 The set  f 0 (t ), f1 (t ),..............f N 1  is an orthonormal basis of
the signal X iff

N 1
 x j (t )  X ,  ai   x j (t )  ai f i (t )
i 0

25
Signal Space Concepts …
 Let x(t) be a deterministic, real valued signal with finite
energy 
Es 

 x 2 ( t )dt

 And let the set  f n (t ), n  0, 1, 2,...............,N 1 be an orthonormal

set of signals or waveforms
 We can approximate x(t) by
 N 1
x(t)  xn 0
n f n (t), where

xn   x(t) f

n (t), n  1, 2,............N  1

26
Signal Space Concepts …
 The set of waveforms  f n (t ), n  0,1, 2,...............N  1 is said to
be complete iff the error energy is zero, i.e.,
 2
N 1
 

 e    x( t )  x ( t ) dt  Es   xn2  0
  n 0

27
Signal Space Concepts …
 Consider the signal x(t) given by
1  0.5  t  0.5
x(t)  
0 elsewhere

 We approximate x(t) by
K
x(t)   xn cos nπ t 1 t  1
n 0

28
Signal Space Concepts …

29
Signal Space Concepts …
 In general, assume that we have a set of waveforms
(signals)
s0 (t ), s1 (t ),..................s M 1 (t )
 And we wish to construct a set of orthonormal waveforms
f 0 (t ), f 1 (t ),..............f N 1 
from the original signal set
 This can be done either:
 Formally by using the Gram-Schmidt orthogonalization
procedure (READ Page 163 of the text); or
 In simple cases, by inspection

30
Signal Space Concepts …
 Example: Find an orthonormal basis for the following set
of waveforms and determine the coordinates of each
waveform in the signal space defined by the orthonormal
basis functions
x1(t) x2(t)

x3(t)

31
Signal Space Concepts …
 By inspection the following orthonormal function can be
the basis for the representation of the three waveforms
shown above

1 1
x1( t )  f1( t ); x2 ( t )  f1( t )  f2( t )  f3( t )
2 2
1 1
x3 ( t )  f1( t )  f2( t )  f3( t )
2 2
In vector form
 1   1 1 
X s  1,0 ,0;
1
X 2  1, ,  ; X 3  1, ,
 2 2  2 2 

32
Chapter 2: Characterization of Communication
Signals and Systems
Overview
 Pulse amplitude modulation
 Phase modulation
 Quadrature amplitude modulation
 Effect of noise

34
Linearly Modulated Digital Signals
 Modulation is the process of encoding information from a
message source in a manner suitable for transmission
 May be done by varying amplitude, phase, frequency, or a
combination of a carrier signal in accordance with the
amplitude of the message (or modulating) signal
 Modulation translates a base-band message signal to a
band-pass signal
 Analogue and digital modulation (this chapter focuses digital
modulation techniques)

35
Linearly Modulated Digital Signals
 Linear digitally modulated signals are expanded in terms of
two orthonormal basis functions of the form
2 2
f1 (t )  cos 2f c t and f 2 (t )  sin 2f c t
T T
 If the low-frequency representation is desired

Slm  xl (t )  jyl (t ) then S m (t )  xl (t ) f1 (t )  yl (t ) f 2 (t )

where xl (t ) and yl (t ) represents the modulating signal

 Modulator maps blocks of k=log2M binary digits at a time

from the information sequence {ak} and selecting one of
M=2k deterministic and finite energy waveform
{Sm(t), m=1,2,…..M} for transmission over the channel

36
Memoryless Modulation – PAM Signals
 Assume that the sequence of binary digits at the input of
the modulator occurs at the rate of R bits/s
 In pulse-amplitude modulation (PAM) signals

 
S m (t)  Re Am g(t) e j 2 πf ct  Am g(t) cos 2 πf ct m  1,2,...M

 0  t  T and {Am, 1  m  M} denotes the M possible

amplitudes corresponding to M=2k possible k-bit blocks or
symbols
 Am takes discrete values or levels, Am = (2m-1-M)d where
2d is the distance between two adjacent signal amplitudes
 g(t) is a real valued signal pulse whose shape influences
the spectrum of the transmitted signal

37
Memoryless Modulation – PAM Signals …
 Symbol rate for PAM signals is R/k, the rate at which
changes occur in the amplitude of the carrier
 Bit interval Tb = 1/R and Symbol interval T = k/R = kTb
 PAM signals have energies
T T
1 1
m   S (t ) dt  Am2  g 2 (t ) dt  Am2  g
2
m
0
2 0
2
 g  Energy of the pulse g(t)

38
Memoryless Modulation – PAM Signals …
 Note that these signals are one dimensional (N=1) and
hence can be represented by the general form
S (t )  S m f (t )

2 g
 Where f(t)   g (t ) cos 2f ct and S m  Am m  1,2,.......M
g 2

39
Memoryless Modulation – PAM Signals …

Signal space diagram for digital PAM for M=2, 4, and 8

40
Memoryless Modulation – PAM Signals …
 The preferred assignment of k information bits to the M=2k
possible signal amplitudes is one in which the adjacent
signal amplitudes differ by only one binary digit (Gray
Encoding)
 Note the Euclidean distance between any pair of signal
points is
1
d e
mn  (Sm  S n ) 
2
ε g Am  An  d 2 ε g m n
2
 The minimum distance between a pair of adjacent signal
point occurs when
m  n  1; e
d min d 2 g

41
Memoryless Modulation – PAM Signals …
 For transmission over channels that does not require
carrier modulation
S m (t )  Am g (t ), m  1,2,...,M
 This is called baseband signal

42
Overview
 Pulse amplitude modulation
 Phase modulation
 Quadrature amplitude modulation
 Effect of noise

43
Memoryless Modulation – Phase Modulated Signals
 In digital phase modulation, the M signal waveform are
 j 2 (
m 1
) 
S m (t )  Re  g (t ) e M
e j 2f ct ; m  1,2,...M .
 
 2 
 g (t ) cos2f ct  (m  1); 0  t  T .
 M 
2 2
 g (t ) cos (m  1) cos 2f c t  g (t ) sin (m  1) sin 2f ct
M M
2
 Where g (t )  signal pulse and  m  (m  1) are the M possible
M
phases of the carrier T
 These signal waveforms have equal energy given by
T T
1 1
   S m2 (t )dt   g 2 (t ) dt   g
0
20 2

44
Phase Modulated Signals …
 PM signal is also represented as a linear combination of
two orthonormal waveforms f1(t) & f2(t) such that

S m (t )  S m1 f1 (t )  S m 2 f 2 (t ); S m  S m1 Sm2 
_

2
Where f1 (t )  g (t ) cos 2f c t and
g
2
f 2 (t )  g (t ) sin 2f c t
g

 Alternatively, these two dimensional vectors may be

expressed as
 g 2 g 2 
S m  S m1 Sm2   
_
cos (m  1) sin (m  1)
 2 M 2 M 

45
Phase Modulated Signals …
 The phase of the carrier signal is used for modulation
(carrying information)
 Every symbol (k bits) is mapped into a given phase
 The total phase is divided equally among all possible
symbols
 The signal space is two dimensional with signals having as
coordinates
 g 2 g 2 
S m  S m1 S m 2   
_
cos (m  1) sin (m  1)
 2 M 2 M 
m  1,2,...................M

46
Phase Modulated Signals …

Signal space diagrams for PSK signals

47
Phase Modulated Signals …
 Example: 8-PSK

48
Overview
 Pulse amplitude modulation
 Phase modulation
 Quadrature amplitude modulation
 Effect of noise

49
Quadrature Amplitude Modulation (QAM)
 Bandwidth efficiency can be obtained by simultaneously
impressing two separate k-bit symbols from the
information sequence {an} on the amplitude of the two
quadrature carriers cos2πfct and sin2πfct such that


Sm (t )  Re ( Amc  jAms ) g (t ) e j 2f c t
, m  1,2,......,M
 Amc g (t ) cos 2f ct  Ams g (t ) sin 2f ct

 Amc and Ams are the information bearing signal amplitudes of the
quadrature carriers
 g(t) is the signal pulse

50
Quadrature Amplitude Modulation …
 Alternatively, QAM signal waveform is represented by
S m (t )  Vm g (t ) cos (2f ct   m )
 Where
 Ams 
Vm  A A
2
mc
2
ms and  m  tan  1

 Anc 

 Note: This may be viewed as a combined amplitude and

phase modulation
 If we take M1 as the PAM levels M2 as the phases, we can
construct an M=M1M2 combined PAM-PSK signal
constellation

51
Quadrature Amplitude Modulation …
 M1 = 2n and M2 = 2m PAM-PSK signal constellation results
in the simultaneous transmission of m+n = log2M1M2 binary
digits occurring at the symbol rate of R/(m+n)
 Examples of combined PAM-PSK signal space diagrams

52
Quadrature Amplitude Modulation …
 Like PSK signals, QAM signals may also be represented
as a linear combination of two orthonormal signal
waveforms f1(t) and f2(t) such that
S m (t )  S m1 f1 (t )  S m 2 f 2 (t );
2 2
Where f1 (t )  g (t ) cos 2f c t and f 2 (t )  g (t ) sin 2f c t
g g
 g g 
S m  S m1 S m 2    Amc
_
Ams ;
 2 2 

 And the Euclidean distance between any pair of signal

vectors is given by
1
 g  2
d (e)
mn  S m  S n   (( Amc  Anc ) 2  ( Ams  Ans ) 2 )
2 

53
Modulation Schemes in Current Systems

54
Overview
 Pulse amplitude modulation
 Phase modulation
 Quadrature amplitude modulation
 Effect of noise
 Power and bandwidth efficiency

55
BPSK
 Binary phase shift keying (BPSK)
 Sends 1 bit/symbol
 Phase change of 1800 per change in bit

56
M-ary PSK
 Example: 8-PSK that sends 3 bits/symbol

 To demodulate 8-PSK, the receiver must determine the

phase within 450

57
M-ary PSK …
 Example: 16-PSK: We could further increase to 4
bits/symbol using

 To demodulate 16-PSK, the receiver must determine the

phase within 11.250

58
Noise Effect
 Like all transmissions, the received signal will be degraded
by noise

 BPSK receiver makes a decision to determine the phase of a

received signal to determine the corresponding binary signal

59
Noise Effect …
 Consider the same noise for 8-PSK

60
Overview
 Pulse amplitude modulation
 Phase modulation
 Quadrature amplitude modulation
 Effect of noise
 Power and bandwidth efficiency

61
Digital Modulation – Factors Influencing Choice
 Power and energy efficiencies
 Provides low BER at low received SNR

 Bandwidth efficiency
 Occupy a minimum bandwidth
 Easy and cost-effective to implement
 Depending on the demands of the particular application,
tradeoffs are made when selecting a digital modulation
 Performance measure for a modulation scheme include
 Power efficiency
 Bandwidth efficiency

62
Power (Energy) Efficiency, P
 Describes the ability of a modulation technique to preserve
the fidelity of the digital message at low power levels
 Fidelity: an acceptable bit error probability

 Often expressed as E N required at the receiver input for a

b
0

certain probability of error (say 10-5)

 where
Eb: signal energy per bit and
N0: noise power spectral density
 Power efficiency measures how favorably the tradeoff
between fidelity and signal power is made

63
Bandwidth Efficiency, B
 Describes the ability of a modulation scheme to
accommodate data within a limited bandwidth
 Measured in bps/Hz
 Sometimes called spectral efficiency
 Defined as the ratio of the throughput data rate per Hertz in
a given bandwidth, i.e.,
data rate
R (bps)
B 
B (Hz) bandwidth
 It reflects how efficiently the allocated bandwidth is utilized
 System capacity of a digital mobile communication
system is directly related to the bandwidth efficiency of
the modulation scheme

64
Bandwidth Efficiency …
 A fundamental upper bound on achievable bandwidth
efficiency is stated by Shannon’s theorem
 The theorem states: “For an arbitrarily small probability of
error, the maximum possible bandwidth efficiency is limited by
the noise in the channel”
 This capacity is given by
C S
ηBmax   log 2 (1  )
B N
 where
C is the channel capacity in bps,
B is the RF bandwidth, and
S/N is the signal-to-noise ratio

65
Bandwidth Efficiency …
 Example: If the SNR of a wireless communication link is 20
dB and the RF bandwidth is 30 kHz, determine the
maximum theoretical data rate that can be transmitted.
 Solution:
SNR=20 dB=100, Bandwidth B=30000 Hz, hence
C = Blog2 (1+S/N) = 30000log2(1+100) = 199.75 kbps

66
Bandwidth Efficiency - GSM System
 Example: What is the theoretical maximum data rate that
can be supported in a 200 kHz channel for SNR=10 dB
and 30 dB? How does this compare to the GSM standard?

 Solution: SNR=10 dB=10, B=200 kHz,

C = Blog2 (1+S/N) = 200000log2(1+10) = 691.886 kbps
 GSM data rate is 270.833 kbps, which is about 40 % of the
theoretical limit for 10 dB SNR conditions
 Bandwidth efficieny = 270/200 =1.35bits/sec/Hz

 For SNR=30 dB=1000, B=200 kHz,

C = Blog2 (1+S/N) = 200000log2(1+1000) = 1.99 Mbps

67
Digital Modulation – Advantages
 Greater noise immunity and robustness to channel
impairments
 Perform well in multipath and fading conditions

 Easier multiplexing of various forms of information (voice,

data, video)
 Greater security
 Allow signal conditioning and processing techniques
 Error-control codes: Detect and/or correct transmission errors
 Equalization: to improve the performance of the overall
communication link
 Source coding
 Encryption
 Possible to be implemented completely in software
68
Some Benefits of Modulation
1. Easy radiation
 Antennas operate effectively when antenna size is comparable
to the wavelength 

2. Frequency matching
 Modulation shifts the spectral of a message signal so as to fit
the frequency band of the channel

3. Multiplexing
 Accommodation for simultaneous transmission of several
baseband signals

69
Chapter 2: Characterization of
Communication Signals and Systems
Spectral Characteristics of Digitally Modulated Signals
 Generally, the available channel bandwidth is limited
 In the selection of the modulation methods, spectral
content of digitally modulated signals be determined
 This helps to take the effect of BW constraint into account

 A digitally modulated signal is a stochastic process since

the information sequences are random
 Need to determined power spectral density (PSD) of these
processes
 From PSD one can find the channel bandwidth required to
transmit the information-bearing signals

71
Spectral Characteristics of …
 Consider a linearly modulated band-pass signal given by

s(t)  Re v(t)e j 2 πf c t

 Where (t) is the equivalent low-pass signal

72
Spectral Characteristics of …
 Autocorrelation function of s(t) is

ss (  )  Re vv (  ) e j 2 πf τ
c

 And its Fourier transform yields the desired expression for
the PSD ss(f) as
Φss (f) 
1
Φvv (f  f c )  Φvv (  f  f c )
2
 Where vv(f) is the PSD of (t)
 To obtain the spectral characteristics of the bandpass
signal s(t), it suffices to determine the autocorrelation
function and power spectral density of the equivalent low-
pass signal (t)

73
Spectral Characteristics of …
 Consider a linear digital modulation method for which (t) is
represented in the general form
n 
v(t)  I
n  
n g(t  nT)

 Where:
 {In} represents the sequence of symbols resulting from mapping
k-bit blocks into corresponding points
 1/T = R/k symbols/s is the transmission rate

 In linear modulation techniques, the amplitude of the

transmitted signal varies linearly with the modulating digital
signal

74
Spectral Characteristics of …
 The autocorrelation function of (t) is
vv (t  τ,t)  E v (t) v(t τ)
1
2
   E I n I m  g  (t  nT) g(t  τ  mT)
1  
2 m n

 The average PSD of (t) is the Fourier transform of its

autocorrelation function, (f)

75
Spectral Characteristics of …
 It is show that the PSD (f) is given by
2
σ μ m  m
2 2
Φvv (f)  G(f)   G T  δ f  T 
i 2 i
2
T T

 Where
 G(f) - Fourier transform of g(t)
 i - Mean of the information sequence
 i2 - Variance of the information sequence
 T - Symbol period
 (f) - Dirac delta function

76
Spectral Characteristics of …

2
σ μ m  m
2 2
Φvv (f)  G(f)   G T  δ f  T 
i 2 i
2
T T

 The first term is a continuous spectrum and its shape

depends on the spectral characteristics of signal pulse g(t)

 The second expression contains discrete frequency

components spaced 1/T apart in frequency
 Each spectral line has power proportional to G(f)2
evaluated at f=m/T

 Note that the power spectral density depends on the

spectral characteristic of the pulse g(t) and the information
sequence {In}

77
Spectral Characteristics of …
2
σ μ m  m
2 2
Φvv (f)  G(f)    T   T 

2
i i
2
G δ f
T T

 Or, the spectral characteristics of (t) can be controlled by

the choice of g(t) and the correlation characteristics of the
information sequence

 If the information sequences has zero mean, i.e. μi =0, the

discrete frequency components will vanish
 This property is most desirable for digital modulation and
can be achieved when the information sequences are
equally likely & symmetrically positioned in a complex plane

78
Spectral Characteristics of …
 Example 1: Consider g(t) to be a rectangular pulse as
shown in the figure below with Fourier transform G(f)

 Rectangular pulse and its energy density spectrum

2
sin πfT  sin πfT 
G(f)  AT and G(f)  (AT) 
2 2

πfT  πfT 

79
Spectral Characteristics of …
 It contains zeros at multiples of 1/T in frequency and it also
decays inversely as the square of the frequency variable
 As a result, all but one of the discrete spectral components
in (f) vanishes
 Thus, upon substitution for G(f)from above, we get
2
 sin πfT 
Φvv (f)  σ i2 A2T    A2 μi2 δ(f)
 πfT 

80
Spectral Characteristics of Digitally …
 Example 2: Consider the case where g(t) is a raised cosine
pulse
A 2 T 
g (t )  1  cos (t  ) for 0  t  T
2 T 2 

Raised cosine pulse and its energy density spectrum

81
Spectral Characteristics of Digitally …
 Its Fourier transform is given as

AT sin πfT  jπ fT
G(f)  e
2 πfT(1 f T )
2 2

 Note the spectrum has zeros at f = n/T; n= +2, +3, +4…….

 Hence, all the spectral components, except those at zero
and f =+1/T vanish
 Compared to that of the rectangular pulse, the spectrum
has a broader main lobe but the tails decay inversely as f6

82