Response of Linear time-invariant system to WSS input

In many applications, physical systems are modeled as linear time invariant (LTI)
systems. The dynamic behavior of an LTI system to deterministic inputs is described by
linear differential equations. We are familiar with time and transform domain (such as
Laplace transform and Fourier transform) techniques to solve these equations. In this
lecture, we develop the technique to analyze the response of an LTI system to WSS
random process.

The purpose of this study is two-folds:

(1) Analysis of the response of a system

(2) Finding an LTI system that can optionally estimate an unobserved random process
from an observed process. The observed random process is statistically related to
the unobserved random process. For example, we may have to find LTI system
(also called a filter) to estimate the signal from the noisy observations.

Basics of Linear Time Invariant Systems:

A system is modeled by a transformation T that maps an input signal x(t ) to an output

signal y(t). We can thus write,

y (t )  T  x(t )

Linear system
The system is called linear if superposition applies: the weighted sum of inputs results in
the weighted sum of the corresponding outputs. Thus for a linear system

T  a1 x1  t   a2 x2  t    a1T  x1  t    a2T  x2  t  

Example : Consider the output of a linear differentiator, given by

d x (t )
y (t ) 
dx
d
Then,
dt
 a1 x1 (t )  a2 x2 (t ) 
d d
 a1 x1 ( t )  a2 x2 (t )
dt dt

Hence the linear differentiator is a linear system.

Linear time-invariant system

Consider a linear system with y(t) = T x(t). The system is called time-invariant if
T x  t  t 0   y  t  t0   t0
It is easy to check that that the differentiator in the above example is a linear time-
invariant system.

Causal system
The system is called causal if the output of the system at t  t0 depends only on the
present and past values of input. Thus for a causal system

y  t0   T  x(t ), t  t0 
Response of a linear time-invariant system to deterministic input
A linear system can be characterised by its impulse response h(t )  T  (t ) where  (t ) is
the Dirac delta function.
x(n) LTI
system

Recall that any function x(t) can be represented in terms of the Dirac delta function as
follows

x(t )   x( )   t  s 

ds

If x(t) is input to the linear system y(t) = T x(t), then


y (t )  T  x( s)   t  s 

ds


  x( s) T   t  s  ds  Using the linearity property 


  x( s ) h  t , s 

ds

Where h  t , s   T   t  s  is the response at time t due to the shifted impulse   t  s  .

If the system is time invariant,

h  t, s   h  t  s 
Therefore for a linear time invariant system,

y (t )   x  s  h  t  s  ds 

x(t ) * h(t )

where * denotes the convolution operation.

Thus for a LTI System,

y (t )  x (t ) * h(t )  h(t ) * x(t )

 x(t ) LTI System y (t )

X ( ) LTI System Y ( )

Taking the Fourier transform, we get

Y    H   X  

where H     FT h  t    h( t ) e
 jt
dt is the frequency response of the system


Response of an LTI System to WSS input

Consider an LTI system with impulse response h(t). Suppose { X (t )} is a WSS process
input to the system. The output {Y (t )} of the system is given by
 
Y  t   h  s  X  t  s  ds   h  t  s  X  s  ds
 
where we have assumed that the integrals exist in the mean square (m.s.) sense.

Mean and autocorrelation of the output process  Y  t  

The mean of the output process is given by,


EY  t   E  h  s  X  t  s  ds


  h  s  EX  t  s  ds


  h s

X ds


 X

 h  s  ds
 X H  0

where H (0) is the frequency response H ( ) at 0 frequency (   0 ) given by

  

 h  t e  h  t  dt
 j t
H ( )  0  dt 
  0 

Therefore, the mean of the output process  Y  t   is a constant

The Cross correlation of the input X(t) and the out put Y(t) is given by

E  X  t   Y  t    E X  t    h  s  X  t  s  ds


  h  s

E X  t    X  t  s  ds

  h s

RX    s  ds

  h  u 

RX    u  du [ Put s   u ]

 h    * RX   

 R XY     h    * R X   
also RYX     R XY     h    * R X   
 h    * RX   
Thus we see that RXY    is a function of lag  only. Therefore, X  t  and Y  t  are
jointly wide-sense stationary.

The autocorrelation function of the output process Y(t) is given by,


 E  Y  t   Y  t    E  h s X  t    s  dsY (t )




 h  s E X  t    s  Y (t ) ds




 h s R XY    s  ds

 h( ) * R XY     h( ) * h(  ) * R X   

Thus the autocorrelation of the output process  Y  t   depends on the time-lag  , i.e.,
EY  t  Y  t     RY    .
Thus
RY     RX    * h    * h   

The above analysis indicates that for an LTI system with WSS input
(1) the output is WSS and
 (2) the input and output are jointly WSS.

The average power of the output process  Y  t   is given by

PY  RY  0 
 RX  0  * h  0  * h  0 
Power spectrum of the output process
Using the property of Fourier transform, we get the power spectral density of the output
process given by

SY     S X    H    H *   
 SX    H   
2

Also note that

R XY     h    * R X   
and RYX     h    * RX   
Taking the Fourier transform of RXY    we get the cross power spectral density S XY   
given by

S XY     H *    S X   
and
SYX     H    S X   

S XY ( ) SY ( ) R ( ) RY ( )
S X ( ) RX ( ) h( ) XY
H ( )
*
H ( ) h( )
Example:

N0
(a) White noise process X  t  with power spectral density is input to an ideal low
2
pass filter of band-width B. Find the PSD and autocorrelation function of the output
process.

H ( )
N
2

B B 

N0
The input process X  t  is white noise with power spectral density S X     .
2

The output power spectral density SY    is given by,

SY     H    S X   
2

N0
 1 B   B
2
N0
  B   B
2

 RY     Inverse Fourier transform of SY   

N sin B
B
1 N 0 j

2 
B
2
e d  0
2 

The output PSD SY    and the output autocorrelation function RY    are illustrated in
Fig. below.
 N0 SY ( )
2

B
O

B 

Example 2:

A random voltage modeled by a white noise process X  t  with power spectral density
N0
is input to an RC network shown in the fig.
2

Find (a) output PSD SY   

(b) output auto correlation function RY   
(c) average output power EY  t  R
2
The frequency response of the system is given by

1
jC
H   
1
R
jC
1

jRC  1

Therefore,
(a)
SY     H    S X   
2

1
 SX   
R C  2 1
2 2

1 N0
 2 2 2
R C  1 2
(b) Taking inverse Fourier transform


N 0  RC
RY     e
4 RC

(c) Average output power

N0
EY 2  t   RY  0  
4 RC