Principles of Communications

Principles of Communications
Chapter 1: Random Process
Yong ZENG （曾勇）
Professor (教授), PhD Advisor（博导）
National Mobile Communications Research Laboratory
(移动通信国家重点实验室）
Purple Mountain Laboratories
(紫金山实验室）

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory)

 Principles of Communications

Chapter 1: Outline
1. Introduction 9. Noise
2. Mathematical Definition of a 10. Narrowband Noise
Random Process 11. Representation of
3. Stationary Processes Narrowband Noise in
4. Mean, Correlation, and Terms of In-phase and
Covariance Functions Quadrature Components
5. Ergodic Processes 12. Representation of
Narrowband Noise in
6. Transmission of a Random Terms of Envelope and
Process Through a Linear Phase Components
Time-Invariant Filter
13. Sine Wave Plus
7. Power Spectral Density Narrowband Noise
8. Gaussian Process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 2

 Principles of Communications

1.1 Introduction
• Two mathematical models:
– Deterministic
– Stochastic (random)
• Received signal in a communication system usually consists of:
– Information-bearing signal
– Random interference
– Channel noise

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 3

 Principles of Communications

Introduction
• Received signal is random in nature
• Impossible to predict the exact signal value in
advance
• But may describe the signal using statistical
parameters:
– Average power
– Power spectral density (PSD)

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 4

 Principles of Communications

Introduction

• Same working
condition
• Recording the noise
waveform at the same
time.

𝑵 completely different output noise waveforms

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 5
 Principles of Communications

1.2 Definition of a Random Process

• Random variable:
– maps the outcome of a random experiment, 𝑠, to a
number, 𝑋(𝑠)
• Random process:
– maps the outcome of a random experiment, 𝑠, to a
time-domain waveform, 𝑋(𝑡, 𝑠)
• Example: Sampling the output of a noisy sensor with:
– A voltmeter random variable
– Oscilloscoperandom process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 6

 Principles of Communications

Definition of a Random Process

• Random processes have two properties:
– They are functions of time
– They are random
• Some concepts:
– Sample point 𝑠𝑗 : each outcome of
the experiment
– Sample space 𝑆: the totality of
sample points

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 7

 Principles of Communications

Definition of a Random Process

• Random process (RP):
𝑋(𝑡, 𝑺) = 𝑋(𝑡)
• For fixed sample point 𝑠𝑗 :
realization of RP:
𝑋(𝑡, 𝑠𝑗) = 𝑥𝑗(𝑡)
• For fixed time 𝑡𝑘 : random
variable 𝑋(𝑡𝑘 )：

 x  t  , x  t  ,   , x  t 
1 k 2 k n k

• The value of a random process at a particular time : random variable

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 8

 Principles of Communications

Ensemble of a Random Process

• Ensemble: All possible time functions that might result from
the random experiment.
Ensemble

Realization/Sample
function 𝑠1

𝑠2

𝑠𝑛

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 9

 Principles of Communications

Random Process Example

• Sinusoidal wave with random phase:
X  t   A cos  2 f ct   

𝐴 and 𝑓𝑐 are constants, and Θ is a random variable:

 1
 , -    
f      2

0, elsewhere

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 10

 Principles of Communications

Random Process Example

• A random process given by
X t   X ,
where 𝑋 is a random variable uniformly distributed in [−1,1]

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 11

 Principles of Communications

CDF and PDF of Random Variable

• For a random variable 𝑋, the cumulative distribution
function (CDF):
FX ( x)  P( X  x)
• The probability density function (pdf):
d
f X ( x)  FX ( x)
dx
• 𝐾-dimensional joint distribution function
FX1 ,..., X k ( x1 ,..., xk )  P[ X 1  x1 ,..., X k  xk ]

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 12

 Principles of Communications

First-Order Distribution of Random Process

• 𝑋(𝑡) at a particular time 𝑡1 is a random variable, characterized
by the first-order CDF and PDF:
𝐹𝑋 𝑡1 (𝑥)

FX ( t1 ) ( x)  P  X (t1 )  x  𝑋(𝑡1)
𝑠1
d
f X ( t1 ) ( x)  FX ( t1 ) ( x)
dx 𝑠2

𝑠𝑛
𝑡1
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 13
 Principles of Communications

Second-Order Distribution of Random Process

• Sampling 𝑋(𝑡) at two time 𝐹𝑋 𝑡 1 ,𝑋 𝑡 2 (𝑥1, 𝑥2)
instances 𝑡1 and 𝑡2 results in two
𝑋(𝑡1) 𝑋(𝑡2)
random variables 𝑋(𝑡1) and 𝑋(𝑡2)
• Can be characterized by the joint 𝑠1
CDF and PDF (second-order
distribution) 𝑠2
FX ( t1 ), X t2  ( x1 , x2 )  P  X (t1 )  x1 , X  t2   x2 

 2 FX ( t1 ), X t2  ( x1 , x2 ) 𝑠𝑛
f X ( t1 ), X ( t2 ) ( x1 , x2 ) 
x1x2 𝑡1 𝑡2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 14

 Principles of Communications

1.3 Strictly Stationary Processes

• A random process that has time-invariant statistical
properties (pdf)
• Strictly stationary process 𝑋(𝑡) :
FX t1  ,, X tk    x1 , , xk   FX t1 ,, X tk   x1 , , xk 

– 𝑋(𝑡𝑘 ): random variable obtained at times 𝑡𝑘

– For all time shifts 𝜏,all 𝑘
– For all possible choices of observation times 𝑡1… 𝑡𝑘

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 15

 Principles of Communications

Strictly Stationary Processes

• The finite-dimensional distributions depend on the
relative time separation between random variables but
not on their absolute time.
• The random process has the same probabilistic
behavior through all time.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 16

 Principles of Communications

Two Special Situations for Stationary Process

FX t1  ,, X tk    x1 ,  , xk   FX t1 ,, X tk   x1 ,  , xk 
• For 𝑘 = 1, we have
FX (t ) ( x)  FX (t  ) ( x)  FX ( x) for all t and 
– The first-order distribution function is independent of time.
• For 𝑘 = 2 and  = −𝑡1 , we have
FX (t1 ), X (t2 ) ( x1 , x2 )  FX (0), X (t2  t1 ) ( x1 , x2 ) for all t1 and t 2

– The second-order distribution function depends on the time

difference between the observation times and not on the
particular times at which the random process is observed.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 17

 Principles of Communications

Strictly Stationary Processes

A  ai  X ti   bi , i  1, 2, 3

Figure Illustrating the probability of a joint event.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 18

 Principles of Communications

Strictly Stationary Processes

Example 1.1
Three spatial windows located at times 𝑡1, 𝑡2, and 𝑡3, the
probability of the joint event:

A  ai  X ti   bi , i  1, 2, 3

In terms of the joint distribution function, this

probability equals:
???
P  A  FX t1 , X t2 , X t3   b1, b2 , b3   FX t1 , X t2 , X t3   a1 , a2 , a3 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 19

 Principles of Communications

Strictly Stationary Processes

Figure Illustrating the concept of stationary in Example 1.1.

• Probability of the joint event 𝐴 is unchanged with time-

shifted window
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 20
 Principles of Communications

1.4 Mean, Correlation and Covariance Function

• Mean (expectation, or ensemble average) of 𝑋(𝑡):


 X (t )  E[ X (t )]   xf X (t ) ( x)dx

𝑓𝑋(𝑡) (𝑥): first-order pdf

• For a strictly stationary process, since 𝑓𝑋(𝑡) (𝑥) is

independent of time 𝑡, the mean is a constant:
 X (t )   X , for all t

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 21

 Principles of Communications

1.4 Mean, Correlation and Covariance Function

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 22

 Principles of Communications

Random Process Example

• Sinusoidal wave with random phase:
X  t   A cos  2 f ct   
𝐴 and 𝑓𝑐 are constants, and Θ is a random variable:
 1
 , -    
f      2

0, elsewhere

Mean:

E[ X (t )]   xf

X (t ) ( x)dx


1

2  A

cos (2f 0t   )d  0 Sample functions of the
random process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 23

 Principles of Communications

Autocorrelation Function
• Autocorrelation of a random process 𝑋(𝑡):
RX (t1 , t2 )  E[ X (t1 ) X (t2 )]
 
   xx f
 
1 2 X ( t1 ), X ( t2 ) ( x1 , x2 )dx1dx2

𝑓𝑋(𝑡1 ),𝑋(𝑡2 ) (𝑥1 , 𝑥2 ): second-order pdf

• For strictly stationary random process, 𝑓𝑋(𝑡1 ),𝑋(𝑡2 ) (𝑥1 , 𝑥2 )

depends only on the difference between 𝑡1 and 𝑡2 , thus

RX (t1 , t2 )  RX (t2  t1 ), for all t1 and t2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 24

 Principles of Communications

Autocovariance Function
• Autocovariance function of a random process 𝑋(𝑡):
C X (t1 , t2 )  E  ( X (t1 )   X (t1 ))( X (t2 )   X (t2 )) 
=R(t1 , t2 )   X (t1 )  X (t2 )
• For strictly stationary process:
C X (t1 , t2 )  E  ( X (t1 )   X )( X (t2 )   X ) 
 RX (t2  t1 )   X2

– Autocovariance function depends only on the time

difference

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 25

 Principles of Communications

Wide-sense Stationary
• The mean and autocorrelation function only provide a
partial description of a random process
• The conditions 𝜇𝑋 𝑡 = 𝜇𝑋 and 𝑅𝑋 (𝑡1 , 𝑡2 )= 𝑅𝑋 (𝑡2 −
𝑡1 ) are not sufficient for 𝑋(𝑡) to be strictly stationary
• Random processes satisfying the above conditions are
wide-sense stationary (WSS):
– Mean is a constant, and
– Autocorrelation function depends only on time difference
• We shall simply refer wide-sense stationary as
stationary process
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 26
 Principles of Communications

Wide-sense Stationary
• The class of strictly stationary processes with finite
second-order moments forms a subclass of the class
of all stationary processes.
Random
Process
Stationary
Process

Strictly
Stationary
Process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 27

 Principles of Communications

Class Activity
• Which of the following statement is true?
A. A wide-sense stationary process is also strictly stationary
B. A strictly stationary process is also wide-sense stationary
C. A strictly stationary process with finite first- and second-
order moments is also wide-sense stationary
D. None of the above

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 28

 Principles of Communications

Class Activity
• Let the random process
𝑋 𝑡 = 3𝑡 + 𝐴,
where 𝐴 is a random variable uniformly distributed in
the range (0,1). Then which of the following is true?
A. 𝑋(𝑡) is strictly stationary
B. 𝑋(𝑡) is wide-sense stationary, but not strictly stationary
C. 𝑋(𝑡) is neither wide-sense stationary, nor strictly
stationary
D. None of the above

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 29

 Principles of Communications

Properties of the Autocorrelation Function

• The autocorrelation function of a stationary process:
RX ( )  E[ X (t   ) X (t )]
1. Mean square value:
RX (0)  E[ X 2 (t )] Average power of 𝑿(𝒕)
2. An even function of τ:
RX ( )  RX ( )
RX ( )  E[ X (t ) X (t   )]  RX ( )
3. Maximum magnitude at 𝜏 = 0:
RX ( )  RX (0), for all 
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 30
 Principles of Communications

Properties of the Autocorrelation Function

• Prove RX ( )  RX (0), for all 
E[( X (t   )  X (t ))2 ]  0
 E[( X 2 (t   )]  2 E[ X (t   ) X (t )]  E[ X 2 (t )]  0
 2 Rx (0)  2 Rx ( )  0
  Rx (0)  Rx ( )  Rx (0)

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 31

 Principles of Communications

Properties of the Autocorrelation Function

4. For random process without long-term memory, DC
power of 𝑋 𝑡 :
PDC  E 2 [ X (t )]=RX ()
lim RX ( )  lim E[ X (t   ) X (t )]
   

 E[ X (t   )]E[ X (t )]
 E 2 [ X (t )]
5. AC power of 𝑋 𝑡 :
 2 ( X )  E[ X 2 (t )]  E 2 [ X (t )]
 RX (0)  RX ()

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 32

 Principles of Communications

Class Activity
• Which of the following CANNOT be an
autocorrelation function of a random process?
A. 𝑅𝑋 𝜏 = cos(2𝜋𝑓𝑐 𝜏)
B. 𝑅𝑋 𝜏 =1
C. 𝑅𝑋 𝜏 = 𝜏 2
𝐴2 1 − 𝜏 , 𝜏 <1
D. 𝑅𝑋 𝜏 = ቊ
0, 𝜏 ≥1

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 33

 Principles of Communications

Orthogonal, Uncorrelated and Independent

• Two random variables 𝑋 and 𝑌 are
 orthogonal if E ( XY )  0

 uncorrelated if Cov( X , Y )  0

where Cov( X , Y )  E  X   X Y  Y    E  XY    X Y

 independent if: f X ,Y ( x , y )  f X ( x ) f Y ( y )

• Independenceuncorrelated, but reverse is not true

• If 𝜇𝑋 = 0 or 𝜇𝑌 = 0, uncorrelated if and only if
E ( XY )  0
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 34
 Principles of Communications

Uncorrelated and Statistically Independent

RX ( )  E[ X (t   ) X (t )]

• For zero-mean random process, if 𝑅𝑋 (𝜏)=0, implies

that signals are uncorrelated for observations
separated by 𝜏
• 𝑅𝑋 (𝜏) signifies how rapidly the random process
changes with time

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 35

 Principles of Communications

Properties of the Autocorrelation Function

Decorrelation time 𝝉0 Physical significance:
• The interdependence
R x ( 0 )  R x (0)  1% of two random
variables by
observing a random
process 𝑋(𝑡) at times
τ seconds apart.
• Decorrelation time 𝜏0

Figure Illustrating the autocorrelation functions of slowly and rapidly

fluctuating random processes.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 36

 Principles of Communications

Random Process Example

• A random process given by
X t   X ,
where 𝑋 is a random variable uniformly distributed in −1,1 .
Find the auto-correlation function.
2
1 x 1
RX (t1 , t2 )  E[ X 2 ]   dx 
-1 2 3

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 37

 Principles of Communications

Example 1.2 Sinusoidal Wave with Random Phase

X (t )  A cos  2 f ct    𝐴 and 𝑓𝑐 are constants, and
 1
 , -    
f      2

0, elsewhere

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 38

 Principles of Communications

Example 1.2 (Cont’d)

X  t   A cos  2 f ct   
The autocorrelation function of 𝑋(𝑡) is:
RX    E  X  t    X  t  
 E  A2 cos  2 f ct  2 f c    cos  2 f ct    
A2 A2
 E cos  4 f ct  2 f c      E cos  2 f c  
2 2
A2  1 A2
  cos  4 f c t  2 f c    d  cos  2 f c 
2   2 2
A2
 cos  2 f c 
2
Sinusoidal wave with random phase is stationary
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 39
 Principles of Communications

Example 1.2 (Cont’d)

A2
RX ( )  cos(2f c )
2
uncorrelated
2n  1

4 fc
RX ( )  0
C X ( )  0

Figure 1.5 Autocorrelation function of a sine wave with

random phase.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 40

 Principles of Communications

Example: Mixing of a Random Process with a

Sinusoidal Process of Random Phase
Y (t )  X (t ) cos(2 f ct  )

𝑋(𝑡): stationary process; ϴ: uniform between 0 and 2𝜋,

ϴ independent of 𝑋(𝑡)
RY ( )  E[Y (t   )Y (t )]
 E[ X (t   ) cos(2 f ct  2 f c  ) X (t ) cos(2 f ct  )]

 E[ X (t   ) X (t )]E[cos(2 f ct  2 f c  ) cos(2 f ct  )]

1
 RX ( ) E[cos(2 f c )  cos(4 f ct  2 f c  2)]
2
1
 RX ( ) cos(2 f c )
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 41

 Principles of Communications

Example 1.3 Random Binary Wave

 A, symbol 1 was sent, 0  t  T
x(t )  
 A, symbol 0 was sent, 0  t  T

1 / T ,0  td  T
fT（ t d） 
 0, otherwise
d

Figure 1.6 Sample function of random binary wave.

The values taken on at any time instant t0 are either x(t0)=A or x(t0)= -A
with equal probability.
E[ X (t )]  0

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 42

 Principles of Communications

Example 1.3 (Cont’d)

 2 | |
 A (1  ), | |<T
RX ( )   T

0, | |  T

Figure 1.7 Autocorrelation function of

random binary wave.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 43

 Principles of Communications

Cross-correlation Function
• Cross-correlation of two random processes 𝑋 𝑡 and 𝑌(𝑡):
RXY (t , u )  E[ X (t )Y (u )], RYX (t , u )  E[Y (t ) X (u )]
• Correlation properties in matrix form:
 RX (t , u ) RXY (t , u ) 
R 
 YX
R (t , u ) R Y (t , u ) 
• Jointly stationary case
 RX   RXY   
R        t u
   
  t u

R
 YX R Y 
• Symmetry property

RXY    RYX   

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 44

 Principles of Communications

Orthogonal, Uncorrelated, and Independence

• Two random processes 𝑋(𝑡) and 𝑌(𝑡) are
 orthogonal if:
RXY (t1 , t2 )  0, t1 , t2
 uncorrelated if:
C XY (t1 , t2 )  0, t1 , t2
 independent if:
f X t1 ,Y t2   x, y   f X t1   x  fY t2  ( y ), t1 ,t2 E[ X (t1 )Y (t2 )]=E[ X (t1 )]E[Y (t2 )]

• If two random processes are independent, then they are

uncorrelated; but the reverse it not true
• For Gaussian random processes, independence and
uncorrelated are equivalent

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 45

 Principles of Communications

Example 1.4 Quadrature-Modulated Processes

A pair of quadrature-modulated processes 𝑋1 (𝑡) and

𝑋2 (𝑡) that are related to a stationary process 𝑋(𝑡) as
X 1 (t )  X (t )cos(2 f ct  )
X 2 (t )  X (t )sin(2 f ct  )
where 𝑓𝑐 is a carrier frequency, and ϴ is uniformly
distributed over the interval [0, 2𝜋]. ϴ is independent of
𝑋(𝑡) .
Find the cross-correlation function of 𝑋1 (𝑡) and 𝑋2 (𝑡)

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 46

 Principles of Communications

Example 1.4
X 1 (t )  X (t ) cos(2 f ct  ), X 2 (t )  X (t ) sin(2 f ct  )

R12 ( )  E[ X 1 (t ) X 2 (t   )]
 E[ X (t ) X (t   ) cos(2 f ct  ) sin(2 f ct  2 f c  )]

 E[ X (t ) X (t   )]E[cos(2 f ct  ) sin(2 f ct  2 f c  )]

1
 RX ( ) E[sin(4 f ct  2 f c  2)  sin(2 f c )]
2
1
  RX ( ) sin(2 f c )
2
R12 (0)  E[ X 1 (t ) X 2 (t )]  0

Random variables obtained by simultaneously observing the

quadrature-modulated processes are orthogonal to each other

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 47

 Principles of Communications

1.5 Ergodic Processes

• Ensemble averages (expectations)：
– Averages “across the process”

 X (t )  E[ X (t )]   xf X (t ) ( x)dx


• Time averages (long-term sample averages)：

– Averages “along the process”

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 48

 Principles of Communications

Time Average
• The time average of the sample function 𝑥(𝑡) of a
stationary process 𝑋 𝑡 :
1 T
 x (T ) 
2T 
T
x(t )dt

• Time average is a random variable, whose mean is

1 T 1 T
E[  x (T )] 
2T  T
E[ x(t )]dt 
2T  T
 X dt   X

• Time average 𝜇𝑥 𝑇 is an unbiased estimate of the

ensemble average 𝜇𝑋

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 49

 Principles of Communications

Ergodic Processes
• A process 𝑋(𝑡) is ergodic in the mean if two
conditions are satisfied:
lim  x (T )   X
T 

lim var[ x (T )]  0
T 

1 T
 x (T ) 
2T 
T
x(t )dt

• Provides practical means for the estimation of

ensemble averages by time averages

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 50

 Principles of Communications

Time-Averaged Autocorrelation Function

RX ( )  E[ X (t   ) X (t )]
• Time-averaged autocorrelation function of the sample
function 𝑥(𝑡) of a stationary process 𝑋(𝑡):
1 T
Rx ( , T ) 
2T T
x(t   ) x(t )dt
• Process 𝑋(𝑡) is ergodic in the autocorrelation function
if
lim Rx ( , T )  RX ( )
T 

lim var[ Rx ( , T )]  0
T 

• For a random process to be ergodic, it has to be stationary;

the converse is not necessarily true
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 51
 Principles of Communications

Class Activity
• True or False?
• If a random process is stationary, then it is also
ergodic.
A. True
B. False

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 52

 Principles of Communications

Cyclostationary Random Processes

A random process X  t  is cyclostationary if its mean and autcorrelation
function are periodic functions with the same period T0 , i.e,
 X  t  T0    X  t 
RX  t1  T0 , t2  T0   RX  t1 , t2 

• Many modulated processes can be modeled as cyclostationary

processes
• We may define the average autocorrelation function as
1
RX    RX  t   , t dt
T0

T0 0

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory)

 Principles of Communications

Example: Mixing of a Random Process with a

Sinusoidal Process
Y (t )  X (t ) cos(2 f ct )
𝑋(𝑡): stationary process, with mean 𝜇𝑋 ;and autocorrelation function 𝑅𝑋 (τ). Is
𝑌(𝑡) a stationary process?
• The mean of Y(t): Y (t )  E[Y (t )]  E[ X (t )cos(2πf ct )]   X cos(2πf ct )
• The autocorrelation of Y(t):

RY (t   , )  E[Y(t   )Y(t )]
 E[X(t   ) X (t ) cos(2πf c (t   )) cos(2πf c t )]
 RX ( ) cos(2πf c (t   )) cos(2πf c t )
1 1 
 RX ( )  cos(2πf c )  cos(4πf c t  2πf c ) 
2 2 
• Both mean and autocorrelation are periodic with period 1/fc
• So 𝑌(𝑡) a cyclostationary random process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 54

 Principles of Communications

Example: Mixing of a Random Process with a

Sinusoidal Process
Y (t )  X (t ) cos(2 f ct )
• The autocorrelation of Y(t):
1 1 
RY (t   , )  RX ( )  cos(2πf c )  cos(4πf c t  2πf c ) 
2 2 
• The average autocorrelation of Y(t):

1 T
RY ( )   RY (t   , )dt 1
T 0 T
fc
RX ( ) T
  cos(2πfc )  cos(4πfct  2πfc )dt
2T 0

1
 RX ( ) cos(2πf c )
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 55

 Principles of Communications

1.6 Transmission of a Random Process

Through a Linear Time-Invariant Filter
• A linear time-invariant (LTI) filter of impulse
response ℎ(𝑡)
• Input: random process 𝑋(𝑡), output: random process
𝑌(𝑡)


Y (t )  X (t )  h(t )   h( 1) X (t  1)d 1


东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 56

 Principles of Communications
Transmission of a Random Process
Through a Linear Time-Invariant Filter

Y (t )  X (t )  h(t )   h( 1) X (t  1)d 1


• Mean of the output process 𝑌(𝑡):


Y (t )  E Y (t )   E   h( 1 ) X (t   1 )d 1 


  
 
  h( 1 ) E  X (t   1 )  d 1   h( 1 )  X (t   1 ) d 1
 

• When 𝑋(𝑡) is stationary, 𝜇𝑋 𝑡 = 𝜇𝑋 , for all 𝑡, so


Y   X  h( 1 ) d 1   X H (0)


Frequency response of the system: H ( f )   h(t )e  j 2 ft dt


𝐻(0): zero-frequency (DC) response of the system

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 57

 Principles of Communications

Autocorrelation Function of the Output


Y (t )  X (t )  h(t )   h( 1) X (t  1)d 1


RY t , u   E Y t Y u 

 E  h  1  X  t   1  d 1  h  2  X  u   2  d 2 
 

   
 
=
  h   h   E  X  t    X  u    d d
1 2 1 2 1 2

 
  h   h   R  t   , u   d d
  1 2 X 1 2 1 2

If 𝑋(𝑡) is stationary, let τ= 𝑡 − 𝑢:

 
RY ( )    h( 1)h( 2) RX (   1   2) d 1d 2
 

 RX ( )*h( )*h( )
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 58
 Principles of Communications
Transmission of a Random Process
Through a Linear Time-Invariant Filter

Y   X  h( 1 ) d 1   X H (0)

 
RY ( )    h( 1)h( 2) RX (   1   2) d 1d 2
 

• Conclusion: if the input to a stable linear time-invariant

filter is a stationary process, then the output of the filter is
also a stationary process.

 
 0
E Y t    h h R    1 d 1 d 2
 

2
1 2 X 2
 
RY ( 0 )  E [Y 2 ( t )]

• The mean-square value of the 𝑌(𝑡) is a constant.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 59

 Principles of Communications

1.7 Power Spectral Density

• Signifies power per unit frequency (W/Hz, dBm/Hz)
• For deterministic power signal 𝑥(𝑡), let 𝑥𝑇 𝑡 denote its
truncated segment, and 𝐹𝑇 (𝑓) denote its Fourier
transform 𝑥(𝑡)

xT (t )  FT ( f )
• Power spectral density of 𝑥(𝑡) : 𝑥𝑇 𝑡
2
F (f)
S x ( f )  lim T
T  T
• How about the power spectral density for a stationary random
process?

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 60

 Principles of Communications

1.7 Power Spectral Density

• For stationary random process 𝑋(𝑡)
– Random sample signal 𝑥(𝑡), its truncated segment 𝑥𝑇 𝑡 ,
and Fourier transform 𝐹𝑇 𝑓
2
F (f)
xT (t )  FT ( f ) S x ( f )  lim T
T  T
– Different sample functions have different spectral densities
– Power spectral density of a sample function can not be used
as the power spectral density of the process
– The power spectral density of the random process should
be considered as the statistical average of the power
spectrum of all samples

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 61

 Principles of Communications

1.7 Power Spectral Density

• Power spectral density of a sample signal 𝑥 𝑡 :
2
F (f)
S x ( f )  lim T
T  T
• Power spectral density of the random process 𝑋 𝑡 :
2
E FT ( f )
S X ( f )  E[ S x ( f )]  lim
T  T

2
1
T /2 T /2
E FT ( f )
 
 j 2 ft j 2 ft '
 E x T (t )e dt x T (t '
)e dt '

T  T /2
T T /2 
T /2 T /2
1
T T/2 T/2
 j 2 f ( t t ' )
 R X (t  t '
)e dtdt '

  t t   
' 2 T
E FT ( f )
T  T  X  d
 j 2 f 
  1  R ( ) e
T

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 62

 Principles of Communications

1.7 Power Spectral Density

• Power spectral density of a stationary random process:
2
E FT ( f )
S X ( f )  lim
T  T
T
  
 lim  1  RX ( )e  j 2 f  d
T 
T 
T 


SX ( f )  

RX ( )e  j 2 f  d

• The power spectral density of a stationary random

process is the Fourier transform of the autocorrelation
function

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 63

 Principles of Communications

1.7 Power Spectral Density

• Linear time invariant filter with impulse response ℎ(𝑡) or

frequency response 𝐻(𝑓)

h 1    H  f  exp  j 2f 1 df


• Input: random process 𝑋(𝑡)

• Output: random process 𝑌(𝑡)
• Mean-square value of the output random process:
 
E Y  t      h  1  h  2  RX  2   1  d 1d 2
2
 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 64

 Principles of Communications

1.7 Power Spectral Density

 
E Y  t      h  1  h  2  RX  2   1  d 1d 2
2
 

h  1    H  f  exp  j 2 f  1  df



E Y  t       H  f  exp  j 2 f  1  df  h  2  RX  2   1  d 1d 2
  
2
  
  
  
  h  2   H  f  RX  2   1  exp  j 2 f  1 d 1dfd 2
  

  2 1   
E Y 2
 t    
H  f   h  2  exp  j 2 f  2   RX   exp   j 2 f   d dfd 2
 

  SX ( f )
E Y 2
 t    | H  f  |  RX   exp   j 2 f   d df
2

  
E Y 2 t    | H  f  |2 S X  f df


东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 65

 Principles of Communications

1.7 Power Spectral Density

• Power spectral density of input process 𝑋(𝑡):

S X  f    R X   exp j 2f d


• Then mean-square value of the output process 𝑌(𝑡):

E Y t    | H  f  |2 S X  f df

2


• Example:
1
| f  f c | f
1,
| H f |   2
1
0, | f  f c | f
2

when 𝑓 is small,
E[Y 2 (t )]  ( 2f ) S X ( f c )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 66

 Principles of Communications

Class Activity
• Which of the following is NOT a unit of power
spectral density?
A. dBm/Hz
B. dB/Hz
C. dBW/Hz
D. W/Hz

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 67

 Principles of Communications

Einstein-Wiener–Khintchine Relations

SX ( f )   RX ( )e  j 2 f 
d


FT IFT

RX ( )   SX ( f )e j 2 f  df


• Power spectral density 𝑆𝑋 (𝑓) and autocorrelation

function 𝑅𝑋 (τ) of a stationary process form a Fourier-
transform pair, with τ and 𝑓 as the variables of
interest

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 68

 Principles of Communications

Properties of Power Spectral Density

• Property 1: zero-frequency value of PSD equals total area
under the graph of the autocorrelation function

SX (0)   RX ( ) d


• Property 2: mean-square value equals the total area under the

graph of PSD

E[ X (t )]   SX ( f )df
2


• Property 3: PSD is always nonnegative

SX ( f )  0

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 69

 Principles of Communications

Properties of Power Spectral Density

• Property 4: PSD of a real-valued random process is an even
function of frequency

SX ( f )  SX ( f )

• Property 5: PSD, appropriately normalized, has the properties

usually associated with a probability density function
SX ( f )
pX ( f )  


SX ( f )df

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 70

 Principles of Communications

Example 1: Sinusoidal Wave with Random Phase

X (t )  A cos  2 f ct   
 1
 , -    
𝐴 and 𝑓𝑐 are constants, and f      2

0, elsewhere

• The autocorrelation function of 𝑋(𝑡) :

A2
RX    E  X  t    X  t    cos  2 f c 
2
• The power spectral density of 𝑋(𝑡) :

A2 A 2
RX ( )  cos(2 fc ) 
FT
 SX ( f )  [ ( f  fc)   ( f  fc)]
2 4

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 71

 Principles of Communications

Example 1: Sinusoidal Wave with Random Phase

A2
RX    cos  2 f c 
2

FT

A2
SX ( f )  [ ( f  fc)   ( f  fc)]
4

 A2
Power: P   S X ( f )df 
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 72

 Principles of Communications

Example 2: Random Binary Wave

 2 | |
 A (1  ), | |<T
RX ( )   T
0, | |  T

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 73

 Principles of Communications

Example 2: Random Binary Wave


T 
R X ( 0)   S X ( f )df  A S X ( f )   A (1  ) exp(  j 2f )d
2 2
 T T

S X ( f )  A2Tsinc 2 ( fT )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 74

 Principles of Communications

Example 3: Mixing of a Random Process with a

Sinusoidal Process
Y (t )  X (t ) cos(2 f ct  )

𝑋(𝑡): stationary process; ϴ: uniform between 0 and 2𝜋,

ϴ independent of 𝑋(𝑡)
RY ( )  E[Y (t   )Y (t )]
 E[ X (t   ) cos(2 f ct  2 f c  ) X (t ) cos(2 f ct  )]

 E[ X (t   ) X (t )]E[cos(2 f ct  2 f c  ) cos(2 f ct  )]

1
 RX ( ) E[cos(2 f c )  cos(4 f ct  2 f c  2)]
2
1
 RX ( ) cos(2 f c )
2
1
SY ( f )  [ S X ( f  f c )  S X ( f  f c )]
4

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 75

 Principles of Communications

PSD’s of Input/Output Processes

 
RY ( )    h( 1)h( 2) RX (   1   2) d 1d 2
 


SY ( f )   R ( ) exp( j 2 f  )d

Y

  
    h( )h(
  
1 2 ) RX (   1   2 ) exp( j 2 f  )d 1d 2 d

 1  2  0
SY ( f )  H ( f ) H * ( f ) S X ( f )

2
H ( f )  H ( f )H * ( f )
SY ( f ) | H ( f ) |2 S X ( f )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 76

 Principles of Communications

PSD’s of Input/Output Processes

   
E Y t    | H  f  | S X  f df   SY  f df
2

2


SY ( f ) | H ( f ) |2 S X ( f )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 77

 Principles of Communications
PSD and the Magnitude Spectrum of a Sample
Function
• We wish to relate the PSD 𝑆𝑋 (𝑓) to the spectral properties of a
sample function 𝑥(𝑡) of an ergodic stationary process 𝑋(𝑡)
• For 𝑥(𝑡) to be Fourier transformable, it must be absolutely
integrable, i.e.,


x (t ) dt  ,

which can not be satisfied with infinite duration.

• So we use a truncated segment of 𝑥(𝑡), whose Fourier
transform is

X  f ,T    x  t  exp   j 2 ft  dt
T

T

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 78

 Principles of Communications

PSD and the Magnitude Spectrum

• Since 𝑋(𝑡) is ergodic, we have
1
RX    lim  x  t    x  t  dt
T

T  2T T

• Furthermore, we have the following Fourier-transform pair

(please try the poof)
1
T
1 periodogram
    2
x (t ) x (t ) dt | X ( f , T ) |
2T T
2T

1  1
x  t    x  t  dt   | X  f , T  |2 exp  j 2 f   df
T

2T 
T  2T

1 
RX    lim  | X  f , T  |2 exp  j 2 f   df
T   2T

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 79

 Principles of Communications

PSD and the Magnitude Spectrum

 1
RX    lim  | X  f , T  |2 exp  j 2 f   df
T   2T

• Take expectation on both sides

RX    lim 
1
  2T
 
E | X  f , T  |2 exp j 2f df
T 

 
  lim
1

E | X  f , T  |2  exp j 2f df

 T  2T 
• Since auto-correlation function and PSD are Fourier transform
pairs, we have
S X  f   lim
1

E | X  f , T  |2 
T  2T

1  T
 lim E |  xt  exp j 2ftdt |2 
T  2T
 T 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 80

 Principles of Communications

Cross-Spectral Densities (CSD)

• 𝑋(𝑡) and 𝑌(𝑡) are jointly stationary processes with their cross-
correlation functions 𝑅𝑋𝑌 τ and 𝑅𝑌𝑋 τ . CSD is defined as:

S XY  f    RXY   exp   j 2 f   d


SYX  f    RYX   exp   j 2 f   d


• Using the formula for inverse Fourier transformation


RYX     SYX  f  exp  j 2 f   df

RXY     S XY  f  exp  j 2 f   df 


• Properties
RXY ( )  RYX ( ) S XY  f   S YX  f   S YX

f 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 81

 Principles of Communications

CSD Example
𝑋(𝑡) and 𝑌(𝑡) have zero mean, and they are individually
stationary. Consider the sum random process
Z (t )  X (t )  Y (t )
Determine the power spectral density of 𝑍 𝑡

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 82

 Principles of Communications

CSD Example
Z (t )  X (t )  Y (t )
Sol: The autocorrelation function of 𝑍 𝑡 is
RZ (t , u )  E[ Z (t ) Z (u )]
 E[( X (t )  Y (t ))( X (u )  Y (u ))]
 E[ X (t ) X (u )  Y (t )Y (u )  X (t )Y (u )  Y (t ) X (u )]
 RX (t , u )  RY (t , u )  RYX (t , u )  RXY (t , u )
 tu

X,Y jointly stationary RZ ( )  RX ( )  RY ( )  RXY ( )  RYX ( )

S Z ( f )  S X ( f )  SY ( f )  S XY ( f )  SYX ( f )

X,Y uncorrelated SZ ( f )  S X ( f )  SY ( f )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 83

 Principles of Communications

Class Activity
• True or False?
• Consider a pair of zero-mean random processes 𝑋(𝑡)
and 𝑌(𝑡), which are individually stationary and also
jointly stationary. Let 𝑍(𝑡)=𝑋(𝑡) + 𝑌(𝑡). Denote the power
spectral density of 𝑋(𝑡), 𝑌(𝑡) and 𝑍(𝑡) as 𝑆𝑋 (𝑓), 𝑆𝑌 (𝑓),
and 𝑆𝑍 𝑓 , respectively. Then 𝑆𝑍 𝑓 = 𝑆𝑋 (𝑓)+𝑆𝑌 𝑓 .

A. True
B. False

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 84

 Principles of Communications

CSD Example 2
• Passing two jointly stationary processes through a pair of
linear time-invariant filters
• Find the CSC of their output processes

 
V (t )   h ( )X (t   )d

1 1 1 1 Z (t )   h (
2 2 )Y (u   2 )d 2


东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 85

 Principles of Communications

CSD Example 2
 
V (t )   h ( )X (t   )d

1 1 1 1 Z (t )   h (

2 2 )Y (u   2 )d 2

RVZ (t , u )  E [V (t ) Z (u )]
 
 E [  h1 ( 1 ) X (t   1 )d 1  h2 ( 2 )Y (u   2 )d 2 ]
 
 
   h ( )h (
  
1 1 2 2 ) E [ X (t   1 )Y (u   2 )]d 1d 2

 
   h ( )h (
  
1 1 2 2 ) RXY (t   1 , u   2 )d 1d 2

X,Y jointly stationary  

RVZ ( )    h (

1 1 )h2 ( 2 ) R XY (   1   2 )d 1d 2

SVZ ( f )  H1 ( f ) H 2* ( f ) S XY ( f )
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 86
 Principles of Communications

1.8 Gaussian Process

• A random variable 𝑌 has a Gaussian distribution if its
probability density function has the form
1  ( y  Y ) 2 
fY ( y )  exp   
2 Y  2 2
Y 
𝜇𝑌 is the mean; 𝜎𝑌 2 is the variance
• Normalized Gaussian distribution 𝑁(0,1)

1  y2 
fY ( y )  exp   
2  2 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 87

 Principles of Communications

1.8 Gaussian Process

• Observe a random process 𝑋(𝑡) for an interval [0, 𝑇]
• A linear functional of 𝑋 𝑡 ：
T
Y   g (t ) X (t ) dt
0

𝑔(𝑡): some weighting function

• The process 𝑋(𝑡) is a Gaussian process if every linear
functional of 𝑋(𝑡) is a Gaussian random variable
• Gaussian process has many properties that make analytic
results possible
• Gaussian process is often the appropriate model for many
physical phenomena because of Central Limit Theorem

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 88

 Principles of Communications

Central Limit Theorem

• 𝑋𝑖 , 𝑖 = 1,2, … , 𝑁, a set of independently and identically
distributed (i.i.d.) random variables, i.e.,
– 𝑋𝑖 are statistically independent.
– 𝑋𝑖 have the same probability distribution with mean 𝜇𝑋 and 𝜎𝑋 2
• Let 𝑌𝑖 be the normalized random variables of 𝑋𝑖 :
Yi 
1
X i   X  i  0,1,2,...,N
X

E Yi   0 varYi   1
• Define a new random variable
N
1
VN 
N
Y
i 1
i

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 89

 Principles of Communications

Central Limit Theorem

• 𝑋𝑖 , 𝑖 = 1,2, … , 𝑁, i.i.d. random variables,
N N
1 1 1
VN 
N

i 1
Yi =
N
 X
i 1
i  X 
X

• Central limit theorem: the probability distribution of 𝑉𝑁

approaches a normalized Gaussian distribution 𝑁(0,1) in the
limit as the number of random variables approaches infinity:
N
1
VN 
N
Y
i 1
i approaches  (0,1) as N  

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 90

 Principles of Communications

Properties of a Gaussian Process

• Property 1: If a Gaussian process 𝑋(𝑡) is applied to a stable
linear filter, then the random process 𝑌(𝑡) developed at the
output of the filter is also Gaussian

Y (t )   h(t   ) X ( )d


Proof: To show 𝑌(𝑡) is Gaussian, we must show that any

linear functional of 𝑌(𝑡) is a Gaussian random variable
T 
Z   gY (t )  h(t   ) X ( )d dt
0 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 91

 Principles of Communications

Properties of a Gaussian Process

T 
Z   gY (t )  h(t   ) X ( )d dt
0 


Z   g ( ) X ( )d


T
g ( )   gY (t )h(t   )dt
0

• Since 𝑋(𝑡) is Gaussian process, 𝑍 is a Gaussian random

variable

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 92

 Principles of Communications

Properties of a Gaussian Process

• Property 2:
If 𝑋(𝑡) is Gaussian, then 𝑋(𝑡1 ), 𝑋(𝑡2 ) … 𝑋(𝑡𝑛 ) are jointly
Gaussian for any 𝑛, with their 𝑛 −fold joint pdf completely
determined by the means
 X ti   EX ti , i  1, 2,......n
and the covariance functions

 
C X  t k , ti   E  X  t k    X  t k    X t       ,
i X ti k , i  1, 2,......, n

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 93

 Principles of Communications

Properties of a Gaussian Process

1  1 
f X ( t1 ),..., X ( tn ) ( x1 ,..., xn )  exp   ( x  μ)T Σ -1 ( x  μ) 
(2 ) n/2 Δ1/ 2  2 
μ  mean vector
 [ X (t1 ) ,... X (t n ) ]
Σ  covariance matrix
 {C X (tk , ti )}kn ,i 1
Σ 1  inverse of covariance matrix
Δ  determinant of covariance matrix Σ

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 94

 Principles of Communications

Properties of a Gaussian Process

• Property 3:
If a Gaussian process is stationary, then the process is also
strictly stationary

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 95

 Principles of Communications

Properties of a Gaussian Process

• Property 4:
If the random variables 𝑋(𝑡1 ), 𝑋(𝑡2 ) … 𝑋(𝑡𝑛 ) obtained by
sampling a Gaussian process at times 𝑡1 , 𝑡2 , … , 𝑡𝑛 are
uncorrelated, that is

E[( X (tk )  X (tk ) )( X (ti )  X (ti ) )]  0, ik

then these random variables are statistically independent

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 96

 Principles of Communications

Properties of a Gaussian Process

1  1 
f X (t1 ),..., X ( tn ) ( x1 ,..., xn )  exp   ( x  μ )T
Σ -1
( x  μ ) 
(2 ) n/2 Δ1/2  2 
E[( X (tk )  X (tk ) )( X (ti )  X (ti ) )]  0, ik

 12 
 
  22   i2  E[( X (ti )  E[ X (ti )]) 2 ]

 
 
  n2 

n
f X (x)   f X i ( xi )
i 1

1  ( x   ) 2

f X i ( xi )  exp  
i xi
 
2 i  2 i
2


东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 97

 Principles of Communications

Class Activity
• True or False?
• If a Gaussian process is wide sense stationary, then it
must be ergodic.

A. True
B. False

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 98

 Principles of Communications

1.9 Noise
• Noise: unwanted signals that tend to disturb the transmission
and processing of signals
• External noise
– Atmospheric noise, galactic noise, man-made noise
• Internal noise
– Shot Noise
– Thermal Noise

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 99

 Principles of Communications

Shot Noise
• Shot noise arises in electronic devices such as diodes and
transistors because of the discrete nature of current flow in
these devices
• Number of arrivers in a pre-defined interval follows Poisson
distribution
• It is difficult to describe statistical characterization of the shot-
noise process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 100

 Principles of Communications

Thermal Noise
• Thermal noise refers to the electrical noise arising from the
random motion of electrons in a conductor
• Usually modeled using the Thévenin equivalent circuit or the
Norton equivalent circuit
• Available noise power is 𝑘𝑇𝑓 watts, where 𝑘 = 1.38 ×
10−23 joules per degree Kelvin is Boltzmann’s constant, 𝑇 is
the absolute temperature in degrees Kelvin, and 𝑓 is
bandwidth in Hertz
• Since the number of electrons in a resistor is very large and
their random motions inside the resistor are statistically
independent of each other, the central limit theorem indicates
that thermal noise is Gaussian distributed with zero mean
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 101
 Principles of Communications

White Noise
• White noise: an idealized form of noise whose power spectral
density is independent of operating frequency
N0
SW ( f )  N 0  kTe
2
– The dimension of 𝑁0 is watts per Hertz
– 𝑘 = 1.38 × 10−23 joules per degree Kelvin, Boltzmann’s constant
– 𝑇𝑒 : equivalent noise temperature of the receiver
• Why is it called white noise?
– The adjective white is used in the sense that white light contains
equal amounts of all frequencies within the visible band of
electromagnetic radiation

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 102

 Principles of Communications

White Noise
N0 N0
SW ( f )  RW ( )   ( )
2 2

• Samples at different times on a white noise are uncorrelated, no

matter how closely they are
• If the white noise is also Gaussian, then the two samples are
statistically independent

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 103

 Principles of Communications

White Noise
• Strictly speaking, white noise has infinite average power and,
as such, it is not physically realizable
• As long as the bandwidth of a noise process at the input of a
system is appreciably larger than that of the system itself, we
may model the noise process as white noise

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 104

 Principles of Communications

Example: Ideal Low-Pass Filtered White Noise

 N0 N0B
 , B f  B RN ( )   exp( j 2 f  )df
SN ( f )   2 B 2

0, | f | B  N0 B sin c(2 B )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 105

 Principles of Communications

Band-limited Noise
• If the white noise 𝑤(𝑡) is Gaussian, the band-limited noise
𝑛(𝑡) at the filter output is also Gaussian, each noise sample
has a mean of zero and variance of 𝑁0 𝐵
• If 𝑛(𝑡) is sampled at the rate of 2𝐵 times per second, the
resulting noise samples are uncorrelated and, being Gaussian,
they are statistically independent
• Additive white Gaussian noise (AWGN):

y (t )  s (t )  n(t )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 106

 Principles of Communications

Example: Correlation of White Noise with a

Sinusoidal Wave
Consider the sample function
2 T
w' 
T 0
w(t ) cos(2 f c t )dt

which is the output of a correlator with white Gaussian noise

𝑤(𝑡) and sinusoidal wave with unit energy. With 𝑤(𝑡) having
zero mean, it follows that the correlator output 𝑤′ has zero mean
too.
E[ w ']  0
Find the variance of 𝑤′.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 107

 Principles of Communications

Example: Correlation of White Noise with a

Sinusoidal Wave
2 T
w' 
T 0
w(t ) cos(2 f c t )dt

 2
T
T 
  E    w(t1 ) cos(2 f ct1 ) w(t2 ) cos(2 f ct2 )dt1dt2 
2

T 0 
0

T
2 T
   E[ w(t1 )w(t2 )]cos(2 f ct1 ) cos(2 f ct2 )dt1dt2
T 0 0
T
2 T
   R (t1 , t2 ) cos(2 f ct1 ) cos(2 f ct2 )dt1dt2
T 0 0 W

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 108

 Principles of Communications

Example: Correlation of White Noise with a

Sinusoidal Wave

2 T T
    RW (t1 , t2 ) cos(2 f ct1 ) cos(2 f ct2 ) dt1dt2
2

T 0 0 RW (t1 , t2 ) 
N0
 (t1  t2 )
N0 2 T T 2

2 T 0 0
 (t1  t2 ) cos(2 f ct1 ) cos(2 f ct2 ) dt1dt2

N0 2 T N0
 g (t ) (t )dt  g (0)
2 T 0
 cos 2
(2 f c t ) dt  
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 109

 Principles of Communications

1.10 Narrowband Noise

• Communication receiver usually uses narrowband filter whose
bandwidth is just enough to pass the modulated signal
• The noise process at the output of such a filter is narrowband noise:
bandwidth 2𝐵, centered on frequency 𝑓𝑐
n( t )  r ( t ) cos( 2f c t   ( t )) B f c

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 110

 Principles of Communications

Representations of Narrowband Noise

• Representation in terms of in-phase and quadrature
components
• Representation in terms of envelope and phase
Envelope Phase
n(t )  r (t ) cos(2 f ct   (t ))
 r (t ) cos( (t )) cos(2 f ct )  r (t ) sin( (t )) sin(2 f ct )

 nI (t ) cos(2 f c t )  nQ (t ) sin(2 f ct )

In-phase component quadrature component

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 111

 Principles of Communications

1.11 Representation of Narrowband Noise in terms of In-

Phase and Quadrature Components

n(t )  nI (t ) cos(2fct )  nQ (t ) sin(2fct )

• 𝑛𝐼 (𝑡): the in-phase component of 𝑛(𝑡)
• 𝑛𝑄 (𝑡): the quadrature component of 𝑛(𝑡)
• Both 𝑛𝐼 𝑡 and 𝑛𝑄 (𝑡) are low-pass signals
• They are fully representative of 𝑛(𝑡), except for 𝑓𝑐
• Given 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡), may generate 𝑛(𝑡): synthesizer
• Given 𝑛(𝑡), may extract 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡): analyzer

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 112

 Principles of Communications

Narrowband Noise Analyzer and Synthesizer

(a) Analyzer (b) Synthesizer

n(t )  2cos(2 f ct )  2nI (t )cos 2 (2 f ct )  2nQ (t )sin(2 f ct )cos(2 f ct )
 nI (t )(1  cos(4 f ct ))  nQ (t )sin(4 f ct )
 nI (t )  nI (t )cos(4 f ct )  nQ (t )sin(4 f ct )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 113

 Principles of Communications
Properties of In-Phase and Quadrature Components of
Narrowband Noise
1. The in-phase component 𝑛𝐼 (𝑡) and quadrature component
𝑛𝑄 (𝑡) of narrowband noise 𝑛 𝑡 have zero mean
2. If the narrowband noise 𝑛 𝑡 is Gaussian, then 𝑛𝐼 (𝑡) and
𝑛𝑄 (𝑡) are jointly Gaussian
3. If the narrowband noise 𝑛 𝑡 is stationary, then 𝑛𝐼 (𝑡) and
𝑛𝑄 (𝑡) are jointly stationary

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 114

 Principles of Communications

Properties of In-Phase and Quadrature

Components of Narrowband Noise
4. Both 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡) have the same power spectral density,
which is related to the power spectral density 𝑆𝑁 (𝑓) of the
narrowband noise 𝑛 𝑡 as

 S N ( f  f c )  S N ( f  f c ), B f  B
S N I ( f )  S NQ ( f )  
0, otherwise

Proof: Refer to Example 1.7

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 115

 Principles of Communications

Properties of In-Phase and Quadrature

Components of Narrowband Noise
 S N ( f  f c )  S N ( f  f c ), B f  B
S N I ( f )  S NQ ( f )  
0, otherwise

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 116

 Principles of Communications

Example 1.7: Mixing of a Random Process with a

Sinusoidal Process
Y (t )  X (t ) cos(2 f ct  )

𝑋(𝑡): stationary process; ϴ: uniform between 0 and 2𝜋,

ϴ independent of 𝑋(𝑡)
RY ( )  E[Y (t   )Y (t )]
 E[ X (t   ) cos(2 f ct  2 f c  ) X (t ) cos(2 f ct  )]

 E[ X (t   ) X (t )]E[cos(2 f ct  2 f c  ) cos(2 f ct  )]

1
 RX ( ) E[cos(2 f c )  cos(4 f ct  2 f c  2)]
2
1
 RX ( ) cos(2 f c )
2
1
SY ( f )  [ S X ( f  f c )  S X ( f  f c )]
4

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 117

 Principles of Communications

Properties of In-Phase and Quadrature

Components of Narrowband Noise
5. 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡) have the same variance as 𝑛(𝑡)

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 118

 Principles of Communications

Properties of In-Phase and Quadrature Components of

Narrowband Noise
6. The cross-spectral density of 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡) is purely
imaginary, as

S N I NQ ( f )   S NQ N I ( f )

 j  S N ( f  fc )  S N ( f  f c ) ,  B  f  B



0, otherwise

Proof: Refer to equation 1.78 and Example 1.4

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 119

 Principles of Communications

Example 1.4 Quadrature-Modulated Processes

A pair of quadrature-modulated processes 𝑋1 (𝑡) and
𝑋2 (𝑡) that are related to a stationary process 𝑋(𝑡) as
X 1 (t )  X (t )cos(2 f ct  )
X 2 (t )  X (t )sin(2 f ct  )

The cross-correlation function of 𝑋1 (𝑡) and 𝑋2 (𝑡) is:

R12 ( )  E[ X 1 (t ) X 2 (t   )]
1
  RX ( ) sin(2 f c )
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 120

 Principles of Communications

Properties of In-Phase and Quadrature Components of

Narrowband Noise
7. If 𝑛 𝑡 is Gaussian and its power spectral density is
symmetric about the mid-band frequency 𝑓𝑐 , then
𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡) are statistically independent.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 121

 Principles of Communications

Example：Ideal Band-Pass Filtered White Noise

• A white Gaussian noise of zero mean and power spectral density
𝑁0 /2
• is passed through an ideal band-pass filter, with mid-band frequency
𝑓𝑐 , and bandwidth 2𝐵
• Determine the auto-correlation function of the filtered noise 𝑛(𝑡)
and its in-phase and quadrature components

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 122

 Principles of Communications

Example：Ideal Band-Pass Filtered White Noise

 fc  B
N0 fc  B N
RN ( )   exp( j 2f )df   0
exp( j 2f )df
 fc B 2 fc B 2

 N 0 B sin c( 2 B )[exp( j 2f c )  exp( j 2f c )]

 2 N 0 B sin c( 2 B ) cos(2f c )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 123

 Principles of Communications

Example：Ideal Band-Pass Filtered White Noise

• The power spectral density of in-phase and quadrature components:
 S ( f  f c )  S N ( f  f c ), B f  B
S N I ( f )  S NQ ( f )   N
0, otherwise

 N 0 , B  f  B
S N I ( f )  S NQ ( f )  
 0, elsewhere

• The auto-correlation function is

R N I ( )  R N Q ( )  2N 0 Bsinc(2B )

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 124

 Principles of Communications

1.12 Representation of Narrowband Noise in terms of

Envelope and Phase Components
n(t )  nI (t ) cos(2f c t )  nQ (t ) sin(2f c t )

nt   r t  cos2f c t   t 
1
r (t )  [nI2 (t )  nQ2 (t )] 2
The envelope of 𝑛(𝑡)
 nQ(t ) 
 (t )  tan  1
 The phase of 𝑛(𝑡)
 nI (t ) 
The envelope 𝑟(𝑡) and phase (𝑡) are both sample functions
of low-pass random processes.
东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 125
 Principles of Communications
Representation of Narrowband Noise in terms of
Envelope and Phase Components
• The pdf of envelope 𝑟(𝑡) and phase 𝜓(𝑡) can be obtained from
that of in-phase and quadrature components
• 𝑁𝐼 and 𝑁𝑄 : denote the random variables obtained by observing
the random processes 𝑛𝐼 (𝑡) and 𝑛𝑄 (𝑡) , respectively
• 𝑁𝐼 and 𝑁𝑄 : independent random variables with zero mean and
variance σ2   nI  nQ
2 2
1
f NI , NQ (nI , nQ )  exp   
2 2  2 2 
 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 126

 Principles of Communications
Representation of Narrowband Noise in terms of
Envelope and Phase Components
• The probability that 𝑁𝐼 lies between 𝑛𝐼 and 𝑛𝐼 + 𝑑𝑛𝐼 , and that
𝑁𝑄 lies between 𝑛𝑄 and 𝑛𝑄 + 𝑑𝑛𝑄 , is
 nI2  nQ2 
f N I , NQ nI , nQ dnI dnQ 
1
exp  dnI dnQ
2 2  2 
2

Define
nI  r cos
nQ  r sin 

Then

dnI dnQ  r dr d

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 127

 Principles of Communications
Representation of Narrowband Noise in terms of
Envelope and Phase Components
• 𝑅 and Ψ: denote the random variables obtained by observing
the random processes 𝑟(𝑡) and 𝜓(𝑡) , respectively
• The probability of 𝑅 and Ψ lying jointly in the shaded area is
1  r2 
f R , (r , ) dr d  exp   r dr d
2 
2 2
 2 
• The joint pdf of 𝑅 and Ψ is
r  r2 
f R , (r , )  exp  
2 
2 2
 2 
• 𝑓𝑅,Ψ (𝑟, 𝜓) is independent of the angle 𝜓. 𝑅 and Ψ are
statistically independent.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 128

 Principles of Communications

Representation of Narrowband Noise in terms of

Envelope and Phase Components

r  r2 
f R , (r , )  exp  
2 
2 2
 2 


f ( )   f R , (r , )dr
0

1  r  r2 

2 0  2
exp   2 dr
 2 
1
 , 0    2 (uniform distribution)
2

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 129

 Principles of Communications
Representation of Narrowband Noise in terms of
Envelope and Phase Components

f R (r )   f R , (r , )d


2 r  r2 
 exp  2 d
0 2
 2 
2

r  r2 
 2 exp  2 , r  0 (Rayleigh distribution)
  2 

Independent of :
f R, (r, )  f ( ) f R (r ) R and  are
independent.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 130

 Principles of Communications
Representation of Narrowband Noise in terms of
Envelope and Phase Components
 r2 
r
f R (r )  2 exp   2  , r0
  2 
• Normalized Rayleigh distribution
r
v
 0.607

fV (v)   f R (r )

  v2 
v exp    , v0
fV (v)    2
0,
 elsewhere

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 131

 Principles of Communications

1.13 Sine Wave Plus Narrowband Noise

x(t )  A cos(2 f ct )  n(t )
 n 'I (t )cos(2 f ct )  nQ (t )sin(2 f ct )
n 'I (t )  A  nI (t )
• Assume 𝑛(𝑡) is Gaussian with zero mean and variance 2
– Both 𝑛𝐼’(𝑡) and 𝑛𝑄 (𝑡) are Gaussian and statistically
independent
– The mean of 𝑛𝐼’(𝑡) is 𝐴 and that of 𝑛𝑄 (𝑡) is zero.
– The variance of both 𝑛𝐼’(𝑡) and 𝑛𝑄(𝑡) is 2.

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 132

 Principles of Communications

Sine Wave Plus Narrowband Noise

1  (n 'I  A)  nQ 
2 2

f N 'I , NQ (n 'I , nQ )  exp   

2 2
 2 2

 nQ (t ) 
r (t )  {[n 'I (t )]  n Q (t )}
2 2 1/ 2  (t )  tan 
1

 I 
n ' (t )

f R , (r , )  ？？

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 133

 Principles of Communications

Sine Wave Plus Narrowband Noise

1  (n 'I  A)  nQ 
2 2

f N 'I , NQ (n 'I , nQ )  exp   

2 2
 2 2

 nQ (t ) 
r (t )  {[n 'I (t )]  n Q (t )}
2 2 1/ 2  (t )  tan 
1

 I 
n ' (t )

r  r 2  A2  2 Ar cos 
f R , (r , )  exp   
2 2
 2 2

• 𝑓𝑅,Ψ (𝑟, 𝜓) cannot be expressed as a product 𝑓𝑅 𝑟 𝑓Ψ 𝜓
• 𝑅 and Ψ are dependent random variables when 𝐴 ≠ 0

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 134

 Principles of Communications

Sine Wave Plus Narrowband Noise

2
f R (r )   f R , (r , )d
0

r  r 2  A2  2  Ar 
 exp    0 exp  2 cos d
2 2
 2 
2
 

1 2

Modified Bessel function of I 0 ( x)  exp( x cos )d
the first kind: 2 0

r r 2  A2   Ar  (Rician
f R (r )  2 exp    I0  2 
  2    
2 distribution)

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 135

 Principles of Communications

Sine Wave Plus Narrowband Noise

• The normalized form:
r A
v a
 
fV (v)   f R (r )
 v2  a2 
fV (v)  v exp    I 0 (av)
 2 

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 136

 Principles of Communications

Sine Wave Plus Narrowband Noise

• When 𝑎 = 0, the Rician distribution reduces to the Rayleigh
distribution.
• The envelope distribution is approximately Gaussian in the
vicinity of 𝑣 = 𝑎 when 𝑎 is large, that is, when the sine-wave
amplitude 𝐴 is large compared with σ, the square root of the
average power of the noise 𝑛(𝑡).

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 137

 Principles of Communications

Chapter 1: Random Process

1. Introduction 9. Noise
2. Mathematical Definition of a 10. Narrowband Noise
Random Process 11. Representation of
3. Stationary Processes Narrowband Noise in
4. Mean, Correlation, and Terms of In-phase and
Covariance Functions Quadrature Components
5. Ergodic Processes 12. Representation of
Narrowband Noise in
6. Transmission of a Random Terms of Envelope and
Process Through a Linear Phase Components
Time-Invariant Filter
13. Sine Wave Plus
7. Power Spectral Density Narrowband Noise
8. Gaussian Process

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 138

 Principles of Communications

Coming Next: Continuous-Wave Modulation

2.1 Introduction 2.9 Superheterodyne
2.2 Amplitude Modulation Receiver
2.3 Linear Modulation 2.10 Noise in CW Modulation
Schemes Systems
2.4 Frequency Translation 2.11 Noise in Linear
Receivers Using Coherent
2.5 Frequency-Division Detection
Multiplexing 2.12 Noise in AM Receivers
2.6 Angle Modulation Using Envelope Detection
2.7 Frequency Modulation 2.13 Noise in FM Receivers
2.8 Nonlinear Effects in FM 2.14 Computer Experiments:
Systems Phase Locked Loop

东南大学移动通信国家重点实验室 (National Mobile Communications Research Laboratory) 139