RANDOM VARIABLES AND

STOCHASTIC PROCESS

CH SIVA RAMA KRISHNA

Asst. Professor
Dept. of ECE
 UNIT-IV
STOCHASTIC PROCESSES-SPECTRAL
CHARACTERISTICS

 Power spectral density of processes and properties,

 Bandwidth of power spectral density,
 Wiener-Khintchine relation,
 Cross power spectral density and properties,
 Relation between cross power spectral density and cross
correlation function.

07/16/2024 07:12 AM RANDOM VARIABLES AND STOC 2

HASTIC PROCESS
 A random process is an ensemble of (collection) number of
time varying functions. Further it is nothing but a random
variable with time added.

 A random variable is a real valued function that assigns

numerical values to the outcomes of physical experiment.

 If time is added to a random variable then it is called random

process.

07/16/2024 07:12 AM RANDOM VARIABLES AND STOC 3

HASTIC PROCESS
 In majority of the communication related problems transmission
power and bandwidth plays a significant role.

 Describing the characteristics of random process is very much

needed in the design of communication experiments.

 Autocorrelation function, Cross-Correlation function, and

Covariance are used to describe the statistical properties of random
processes in time domain.

 Similarly frequency domain can also be used to characterize the

random processes. Power spectral density is the fundamental tool
used to calculate the average power of random processes.
 Fourier transform is the fundamental tool used to get the
frequency information of a signal (or) a process.

 However for any process to have Fourier transform must

satisfy the “Dirichlet’s conditions”.

 One’s Fourier transform is applied then the average power is

area under power spectral density curve.
POWER SPECTRAL DENSITY
Let 𝑥(𝑡) be the one sample function of a random process 𝑋(𝑡). Further consider
𝑥𝑇 (𝑡) represent the portion of 𝑥ሺ𝑡ሻ 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 − 𝑇 < 𝑡 < 𝑇.
𝑥ሺ𝑡ሻ − 𝑇 < 𝑡 < 𝑇
𝑥𝑇 ሺ𝑡ሻ = ቄ
0 𝑒𝑙𝑠𝑒𝑤ℎ𝑒𝑟𝑒
The energy contained in the waveform 𝑥𝑇 (𝑡) is in the interval ൫– 𝑇, 𝑇൯is
𝑇 𝑇

𝐸 ሺ𝑇ሻ = න 𝑥𝑇 2 ሺ𝑡ሻdt = න 𝑥 2 (𝑡) dt

−𝑇 −𝑇

By dividing the above expression by 2T, we obtain the power 𝑃(𝑇) of the truncated
(small) portion is
𝑇 𝑇
1 2
1
𝑃ሺ𝑇ሻ = න 𝑥𝑇 ሺ𝑡ሻdt = න 𝑥 2 ሺ𝑡ሻdt
2𝑇 2𝑇
−𝑇 −𝑇
From the above equation it is observed that
 It does not represent the power in an entire sample function.
 The expression is only the power in one sample function and
does not represent entire random process.
Hence to obtain power density spectrum for the random process
 Let making T arbitrarily large 𝑖. 𝑒; 𝑇 → ∞ and
𝑃(𝑇) is actually a random variable with respect to the random
process. By taking the expected value for 𝑃(𝑇), we can obtain an
average power 𝑃𝑋𝑋 for the random process.
𝑃𝑋𝑋 = 𝐸 ሾ𝑃(𝑇)ሿ
 𝑃𝑋𝑋 = 𝐸 ሾ𝑃(𝑇)ሿ
𝑇
1
𝑃𝑋𝑋 = lim න 𝐸 ሾ𝑥 2 (𝑡)ሿdt
𝑇→∞ 2𝑇
−𝑇

1 ∞ 𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑃𝑋𝑋 = න lim 𝑑𝜔
2𝜋 −∞ 𝑇→∞ 2𝑇
𝐸ൣȁ𝑋 𝑇 (𝜔 )ȁ2 ൧
The term lim is called as power spectral density (PSD)
𝑇→∞ 2𝑇

𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
Where 𝑋𝑇 ሺ𝜔ሻ = 𝐹. 𝑇ሾ𝑥𝑇 (𝑡)ሿ
𝑇

= න 𝑥 ሺ𝑡ሻ 𝑒 −𝑗𝜔𝑡 𝑑𝑡
_𝑇
Properties of power spectral density
1. Power spectral density is always a non-negative quantity.
𝑆𝑋𝑋 ሺ𝜔ሻ ≥ 0
Proof:
It is known that Power spectral density of a random process X(t) is
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
It is known that the term ȁ𝑋𝑇 (𝜔)ȁ2 is always positive and expected
value of appositive quantity is always positive. Thus
𝑆𝑋𝑋 ሺ𝜔ሻ ≥ 0
 2. Power spectral density is an even function
𝑆𝑋𝑋 ሺ−𝜔ሻ = 𝑆𝑋𝑋 ሺ𝜔ሻ
Proof:
It is known that Power spectral density of a random process X(t) is
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
It is known that the term ȁ𝑋𝑇 (−𝜔)ȁ2 is always equal to ȁ𝑋𝑇 (𝜔)ȁ2
𝐸 ሾȁ𝑋𝑇 (−𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ−𝜔ሻ = lim
𝑇→∞ 2𝑇
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
= lim
𝑇→∞ 2𝑇
𝑆𝑋𝑋 ሺ−𝜔ሻ = 𝑆𝑋𝑋 ሺ𝜔ሻ
 3. Power spectral density is always real valued function.
i.e; Imaginary part of 𝑆𝑋𝑋 ሺ𝜔ሻ = 0
Proof:
It is known that Power spectral density of a random process X(t) is
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
It is known that the term ȁ𝑋𝑇 (𝜔)ȁ2 is always positive and real
quantity. Thus
Imaginary part of 𝑆𝑋𝑋 ሺ𝜔ሻ = 0
 4. Power spectral density of derivative of random process is equal to 𝜔2 times power spectral
density of random process.
𝑆𝑋ሶ𝑋ሶሺ𝜔ሻ = 𝜔2 𝑆𝑋𝑋 ሺ𝜔ሻ
It is known that Power spectral density of a random process X(t) is
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
𝑆𝑋𝑋 ሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
Where
𝑇

𝑋𝑇 ሺ𝜔ሻ = 𝐹. 𝑇ሾ𝑥 𝑇 (𝑡)ሿ= න 𝑥 ሺ𝑡ሻ 𝑒 −𝑗𝜔𝑡 𝑑𝑡

_𝑇
𝑇
𝑑 𝑑
𝑋ሶ𝑇 ሺ𝜔ሻ = 𝐹. 𝑇 ൤ 𝑥 𝑇 (𝑡)൨= න 𝑥 ሺ𝑡ሻ 𝑒 −𝑗𝜔𝑡 𝑑𝑡
𝑑𝑡 𝑑𝑡
_𝑇

𝑋ሶ𝑇 ሺ𝜔ሻ = ሺ𝑗𝜔ሻ 𝑋𝑇 ሺ𝜔ሻ

2
𝐸 ቂห𝑋ሶ𝑇 (𝜔)ห ቃ
𝑆𝑋ሶ𝑋ሶሺ𝜔ሻ = lim
𝑇→∞ 2𝑇
𝐸 ሾȁሺ𝑗𝜔ሻ 𝑋𝑇 ሺ𝜔ሻȁ2 ሿ
= lim
𝑇→∞ 2𝑇
2
𝐸 ሾȁ𝑋𝑇 (𝜔)ȁ2 ሿ
= 𝜔 lim
𝑇→∞ 2𝑇
𝑆𝑋ሶ𝑋ሶሺ𝜔ሻ = 𝜔2 𝑆𝑋𝑋 ሺ𝜔ሻ
1. The power spectral density and time average of auto correlation function form a Fourier transform
pair.
1 ∞
න 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑒 𝑗𝜔𝜏 𝑑𝜔 = 𝐴ሾ𝑅𝑋𝑋 ሺ𝑡, 𝑡 + 𝜏ሻሿ
2𝜋 −∞
1. Demonstrate whether given power spectral densities are valid or not.

𝜔2 cosሺ3𝜔ሻ ȁωȁ
ሺ𝑖 ሻ𝑆𝑋𝑋 ሺ𝜔ሻ = 6 ሺ𝑖𝑖 ሻ𝑆𝑋𝑋 ሺ𝜔 ሻ = ሺ𝑖𝑖𝑖 ሻ𝑆𝑋𝑋 ሺ𝜔 ሻ =
𝜔 + 3𝜔 2 + 3 1 + 𝜔2 1 + 2𝜔 + 𝜔 2
1. Determine and plot the power density spectrum of random process whose auto
2
A0
correlation function RXX ( )  Cos( 0 ) .
2
Bandwidth of random process:
 Whenever it is required to describe a random variable,
probability density function is used as one attribute/ parameter.

 In this the standard deviation is a measure of the spread from

a given reference value.

 This reference value can be (mean) either zero (or) a non-zero

value.

 The standard deviation indicates how the probabilities vary for

a given random variable.

 Similar to the standard deviation, the power spectral density

also has a parameter that is a measure of spread (or)
distribution of power. This is called as bandwidth.
 Let X(t) be a base band process.
 For a base band process, the spectral components are concentrated over the
origin.
 For such a base band process, the bandwidth can be the second moment upon
normalization.
 The rms bandwidth of a base band process is given as
∞
‫׬‬−∞ 𝜔2 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔
𝜔2 𝑟𝑚𝑠 = ∞
‫׬‬−∞ 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔
Similarly the rms bandwidth can also be defined for band pass processes. For
a band pass process, the spectral/ frequency components are concentrated
over some reference value (or) mean value 𝜔0 .
The mean frequency of a band pass process is given as
∞
‫׬‬0 𝜔2 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔
𝜔
ഥ0 = ∞
‫׬‬0 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔

The rms bandwidth of band pass process is

∞
ഥ0 )2 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔
‫׬‬0 (𝜔 − 𝜔
𝜔2 𝑟𝑚𝑠 = ∞
‫׬‬0 𝑆𝑋𝑋 ሺ𝜔ሻ 𝑑𝜔
1. Calculate the rms bandwidth of a random process whose power
spectral density is given as
𝑠𝑋𝑋 ሺ𝜔ሻ = 𝑃 𝑓𝑜𝑟 ȁ𝜔ȁ ≤ 𝑊
=0 ȁ𝜔ȁ > 𝑊.
1. Determine and plot the power density spectrum of random process whose auto correlation
function
𝟏 − ȁ𝝉ȁ 𝒇𝒐𝒓 ȁ𝝉ȁ < 𝟏
𝑹𝑿𝑿 ሺ𝝉ሻ = ቄ
𝟎 𝑶𝑻𝑯𝑬𝑻𝑾𝑰𝑺𝑬
1. Obtain cross correlation function if cross power spectral density is given
jb
S ( )  a 
XY
, W    W
W
0 , Otherwise
1. Determine and plot the power density spectrum of random process whose auto
correlation function
ȁ𝝉ȁ
𝑨 ቈ𝟏 − ቉ 𝒇𝒐𝒓 − 𝑻 ≤ 𝝉 ≤ 𝑻
𝑹𝑿𝑿 ሺ𝝉ሻ = ቐ 𝟎 𝑻
𝟎 𝑶𝑻𝑯𝑬𝑻𝑾𝑰𝑺𝑬