Linear Time-Invariant Systems (LTI)

Introduction
This section introduces the analysis of Causal, Linear, Time-invariant, Continuous systems.
 Causality: The system response depends only on inputs in the past. Excitations that will occur
in the future can have no effect on the present output.
 Linearity: Superposition applies to these systems.
If x1(t) y1(t)
and x2(t) y2(t)

Then a* x1(t) + b* x2(t) a* y1(t) + b* y2(t)

Nonlinear systems are common in the real world. You will run into them in your studies of
Communication Systems and Digital Signal Processing. Don’t try to apply the methods
described here on non-linear systems, as they will NOT work.
 Time Invariance: These systems will behave the same now as they did yesterday. The phase of
the moon has no effect. Time varying systems are also common, but we will assume an
unchanging world here.
 Continuous: All variables are defined along the real line (or in the complex plane). Time
progresses forward in a smooth continuous manner. System behavior can be completely
described by a set of differential equations.
The input-output relationship of an LTI system can be defined in one of several ways
 Time-Domain Microscopic view: This was the first approach you used to analyze physical
systems. The system behavior is described at each instant of time in terms of a set of
Differential Equations and a present State.
 Time-Domain Macroscopic view: Here the system is modeled as having memory. The system
output is produced as a weighted sum (integral) of its past inputs. The weighting function is
called the system Impulse Response and the output is the Convolution of the past input with the
Impulse response.
 Frequency Domain view: In the section on Fourier Signal Analysis the concept of the
frequency domain was introduced. Now we can use the System Transfer Function to describe
how the system processes the frequency domain representation of the input signal to produce the
frequency domain representation of the output signal. Note that a closely related approach based
on the Complex Frequency Domain (s-Domain) is covered in the section on LaPlace
Transform Analysis. Both approaches are equally valid. Fourier is heavily used in
Communications, while LaPlace is preferred for Feedback Controls and, in a discrete-time
formulation (the Z-Transform), for Digital Signal Processing.

LTI System
Differential Equations yt   xt   ht 
x(t)
Impulse Response – h(t)
X(j), X(s) Transfer Function – H(j), H(s) Y(j) = X(j)*H(j)
Y(s) = X(s)*H(s)
X(j), X(s)
J. N. Denenberg February 4, 2013
LTI Systems Page 2

The Convolution Integral

Figure 1: x(), the Input

2.5

1.5

0.5

0
=t0 =t
11

13

15

17

19

21

23

25

27

29
-5

-3

-1

Assume that the input, x(), to an LTI system started at time t0 (the input was zero for all time prior to t0)
and has continued to the present time, t, as shown below.

Figure 2: x( n), the Approximation to the Input

2.5

1.5

0.5

0
0 N=t
1

11

13

15

17

19

21

23

25

27

29

31

33

35

We can approximate this input as a series of rectangular pulses having the same area under the curve as
shown in Figure 2.

These graphs are given in terms of the variable , the variable t is reserved for the time of observation of
the output signal. The interval from  = t0 to  = t is divided into subintervals of width  each centered
about an value of n = t0 + n*.
Now perform the following experiment. Apply a rectangular pulse of unit strength and width  to the
input of our LTI system. Lets call the resulting output f(t, n).

J. N. Denenberg February 4, 2013

LTI Systems Page 3

 f(t, n) is the output at time t due to a rectangular pulse of unit amplitude and width  that
occurred at time  = n .
The output of the system at time t due to the nth pulse of the approximate input is then the value of the
input at time n ,which is x(n), times f(t, n). Using superposition the total output from the system at
time t is then approximated by the sum:
N
yt    x n  * f t , n 
n 0

Multiplying and dividing each term in the sum by  yields:

N
 1 
yt    x n  *  f t , n  
n 0   
 1 
  f t , n 
1
Note that the term is the output of the time t due to a pulse of amplitude  that occurred
at time  = n . The area of this input pulse ise equal to unity. Our approximation gets better as 
approaches zero so take the limit of y(t) as  0 changing the sum to an integral.
 1 
t
f t , n  
y t    x  * 
 d
 
lim   0
 t 0

The term in the brackets becomes the output at time, t, of the system to (t-), a Dirac Delta function or
impulse that occurred at time  . It is usually denoted as h(t, ), or since our system is time invariant
simply h(t-). This function, h(t), is called the Unit Impulse Response of the system (which happens to
be the Inverse Fourier Transform of the Transfer Function, H(j) ). The output then is given by:
t
yt    x  * ht   d
 t 0

or, where it is up to you to determine the limits on the integral from the nature of the two functions:

yt    x  * ht   d
  

This is known as the Convolution Integral and is denoted as:

yt   xt   ht 
Note: The meaning of Convolution is that an LTI system can be modeled as having a memory that
stores all past input. Acording to this model, the LTI system determines its output by performing
a weighted sum of all past inputs using the Impulse Response as the weighting factor.
Continuous systems seldom actually function this way, but this model accurately determines the
output. Many Discrete-Time LTI systems ARE built according to the Convolution model.
They are called Finite Impulse Response systems since their memory has a limited capacity.

J. N. Denenberg February 4, 2013

Convolution as an Operator
Commutative Law
xt   yt   yt   xt 
Proof:

xt   yt    x  * y t   d
 

Let u = t -  , therefore  = t – u and


xt   yt    xt  u  * yu  du 
u 

Reversing the limits is the same as multiplying by –1

 
xt   yt    xt  u  * yu du   yu  * xt  u du  yt   xt 
u   u  

Q. E. D.

Distributive Law
xt   yt   zt   xt   yt   xt   zt 
Proof:
Integration is a linear operator and multiplication distributes over addition

Associative Law
xt   yt   zt   xt   yt   zt 
Proof:
Due to the convolution Law of Fourier transforms
xt   yt   zt   X ( j ) * Y  j  * Z  j 
But multiplication is Associative so
X ( j ) * Y  j  * Z  j   X  j  * Y  j * Z ( j )
Therefore
xt   yt   zt   X  j  * Y  j * Z  j   xt   yt   zt 
Q. E. D.

J. N. Denenberg February 4, 2013

LTI Systems Page 5

Example Convolutions
h(t) = 1 for 0 < t < 1; 0 elsewhere
1.5

0.5

0
-2 -1 0 1 2 3 4
Time

Convolution Example 1: Simple Rectangular Functions

x(t) = 1 for 0 < t < 2; 0 elsewhere

1.5

0.5

0
-2 -1 0 1 2 3 4
Time

h(-)

1.5

0.5

0
-2 -1 0 1 2 3 4
Time

First flip h(t) by letting t = -

J. N. Denenberg February 4, 2013

LTI Systems Page 6

Case 1: t < 0

1.5

x()
1
h(t-)
0.5

0
-2 -1 0 1 2 3 4
t-1 t

Now shift h(-) to the time for Case 1 by replacing (–) with t - 

Case 2 moves the front edge of h(t-) into x() so the output is the shaded area t

For all of Case 3 h(t-) is fully within x() so the output is 1

J. N. Denenberg February 4, 2013

LTI Systems Page 7

Case 4: 2 < t < 3

1.5

x()
1

h(t-)
0.5

0
-2 -1 0 1 t-1 2 t 3 4

In Case 4 h(t-) is exiting x() so the output is [ 2 – (t-1) ]*1 or (3-t)

Case 5: t > 3

1.5

x()
1
h(t-)
0.5

0
-2 -1 0 1 2 3 4
t-1 t

For all of the last Case t > 3 and there is no overlap so the output is 0


yt   xt   ht    x  * ht   d
So now we can plot the output,  and we are done.

y(t)

1.5

0.5

J. N. Denenberg February 4, 2013

0
-2 -1 0 1 2 3 4
Time
LTI Systems Page 8

x(t) = 1 for 0 < t < 2; 0 elsewhere

1.5

0.5

0
-4 -3 -2 -1 0 1 2 3 4 5

Time

Convolution Example 2: A Triangular Function

h(t) = t for 0 < t < 3; o elsewhere

3.5

2.5

1.5

0.5

0
-4 -3 -2 -1 0 1 2 3 4 5
Time

The input signal x(t) is a unit rectangular pulse from t = -1 to t = 1

The system Impulse Response is triangular
Find the output, yt   xt   ht 

J. N. Denenberg February 4, 2013

LTI Systems Page 9

h(- )

3.5

2.5

1.5

0.5

0
-4 -3 -2 -1 0 1 2 3 4 5
Tim e

Case 1: t < 1

h(t-) 2

x()
1

0
-4 -3 -2 -1 0 1 2 3 4 5
t-1 t

First flip the Impulse Response by substituting t = -

Now slide it back to start at a value of t  -1 and plot the signal on the same chart
There is no overlap so y = 0 for t < -1

J. N. Denenberg February 4, 2013

LTI Systems Page 10

Case 2: 0 < t < 1

3.5

2.5

2
h(t-)
1.5
x()
1

0.5

0
-4 t-3 -3 -2 -1 t 0 1 2 3 4 5

Now slide the tip of h(t-) just past t = -1 to set up Case 2

The shaded area is the integral of the product of the two functions. And:

t
t
 2
yt    t    *1d  t *   
1 1
 t2  t 
  1  2   1 2 2 for 0 < t < 1
Case 3 is set up by sliding t to just past t = 1. Now the complete signal lies within the memory of the
system.

J. N. Denenberg February 4, 2013

LTI Systems Page 11

Now
1
1
 2
y t    t    *1d  t *     2t
  1  2   1 for 1 < t < 2
The fourth Case occurs when the back edge of h(t-) crosses  = -1. This is when t = 2.
1
1
 2
yt    t    *1d  t *   
1
  t2  t  4
Now  t 3  2   t 3 2 for 2 > t > 4

J. N. Denenberg February 4, 2013

LTI Systems Page 12

Case 5: t > 4

3.5

2.5

2
h(t-)
1.5
x()
1

0.5

0
-4 -3 -2 -1 0 1 t-1 2 3 4 t 5

The last Case is when t > 4. Now there is no overlap and the output remains at zero.

y(t)

4.5
Case 4

3.5

2.5 Case 2

2
Case 3 Case 5
1.5
Case 1
1

0.5

0
-2 -1 0 1 2 3 4 5
Time

Now we can plot y(t). Note that it is a continuous function. This is the normal case (the exception is
when there are Impulse Functions in either the signal or h(t) ). Use this fact to check your work by
comparing the values at the boundary conditions between cases.

J. N. Denenberg February 4, 2013

LTI Systems Page 13

Convolution Example 3: The RC Low Pass Filter

The input signal x(t) is a unit rectangular pulse from t = 0 to t = t0
The circuit is:

Find the Impulse Response of the circuit using the Transfer Function:
1 1
jC
H  j    RC
1 1
R  j
jC RC Since it is a simple AC voltage divider
From Example 1 of the section on Fourier Transforms the inverse transform of this Transfer function is:
t
1  RC
ht    * U t 
RC which looks like:
h(t)=[1/RC]*exp(-t/RC)*U(t)

1.5

1/RC1

0.5

0
-1 0 1 2 3 4 5
t=1/RC
-0.5

J. N. Denenberg February 4, 2013

LTI Systems Page 14

vin(t) = 2 for 0 < t < 1; 0

elsewhere

0
-2 0 2 4
-1

Time

The input, vin, is:

Case 1: t < 0

x()
2

1
h(t-)

0
-4 -3 -2 -1 0 1 2 3 4 5
t

Now using Convolution, find the output:

Flipping h(t), sliding it to the left, t < 0, we have Case 1:
And of course y(t) = 0 for t < 0 since there is no overlap.

J. N. Denenberg February 4, 2013

LTI Systems Page 15

Case 2: 0 < t < 1

x( )
2

1
h(t- )
0
Case-42 is while
-3 the leading
-2 edge -1
of h(t-) is0 within the
1 square 2pulse or when
3 0 < t4< 1 5
t
Now the integral becomes:

vout t   xt   ht    x  * ht   d
 

t

t  
vout t  
1

0
 2 * RC  RC
d

t t 

vout t  
2
*  RC   RC d
RC  0

 t

t
 
vout t  
2
* RC
 RC *  
RC
RC    0
 t

t
 RC  
t
 RCt   
t

vout t   2 *  RC
   2 *  RC
  1  2 * 1   RC

   0     for 0 < t < 1

J. N. Denenberg February 4, 2013

LTI Systems Page 16

Case 3: 0 < t < 1

x( )
2

1
h(t- )
0
Case-43 is the final
-3 case-2 -1 for t 0
and it is good >1 1 2 3 4 5
t
Now the integral becomes:
1 t  
1 
vout t    2*  RC
d
0
RC
t 1 

vout t  
2
*  RC   RC d
RC  0

 1

t
 
vout t  
2
*  RC  RC *  RC 
RC    0
 1

t
 RC  
t
 RC
1

vout t   2 *  RC
   2 * 
RC
  1
   0  

J. N. Denenberg February 4, 2013

LTI Systems Page 17

The Output

y(t)
1

0 t
-1 0 1 2 3 4 5

Moreover, we can now plot the output:

Review: We will next solve the same problem using methods from earlier courses.

vin(t) = 2 for 0 < t < 1; 0

elsewhere

0
-2 0 2 4
-1

Time

Using Fourier Transform Analysis (Frequency Domain):

1
H  j   RC
1
 j
Again RC and
Taking the Fourier Transform of vin(t)

   
1
Vin  j    2 *   jt dt  
2  jt 1 2  j 2
   1  1    j
t 0
j t  o j j

J. N. Denenberg February 4, 2013

LTI Systems Page 18

The Fourier Transform of the output is then the product of this result with the Transfer function. Or:
1
Vout  j  
2
j

1    j *
1
 RC
 j
RC
Or,

2
j
Vout  j   RC  
*   j  1
 1 
  j  * 
 RC 
Using a Partial Fraction Expansion,
 
A 
Vout  j    
B
 *   j  1  
  1 
   j  
  RC 
2
j
A RC  j2
 1 
  j 
 RC   0
And,
2
j
B RC 2

j

RC

Or,
   
 j2   
Vout  j    
2
   
 *   j  1   *   j  1   
j2 2
 
*   j  1 
  1      1  j  
   j    
  RC  
 RC  
The first term is the Fourier Transform of the original input signal. Expanding the second term yields:
 
 2  j 
Vout  j   Vin  j   
2
 
  1  j   1  j  
  RC   RC  
We can now take the Inverse Fourier Transform term by term.

J. N. Denenberg February 4, 2013

LTI Systems Page 19

t 1
  RCt  
vout t   vin t   2 * U t   2 RC * U t  1
 
Checking this out against our result using Convolution:
t 1
 t  
vout t   vin t   2 RC * 0  2 RC * 0  vin t   0
Case 1 (t < 0):  
t 1
  RCt   
t
 
t

vout t   vin t   2 *1  2 RC
* 0  vin t   2 RC 
 2 * 1   RC 

Case 2 (0 < t < 1):    
t 1
 t     t 1 
t
 
t
 1 
vout t   vin t   2 RC *1  2 RC *1  0  2 RC   RC   2 RC   RC  1
Case 3 (t > 1):      
We have the same result!

Solve the same problem once more but use Differential Equations (Microscopic Time Domain):
Remember the original circuit diagram

The Kirchoff loop equation is:

vin t   R * it  
it dt
1
C
First solve the Homogeneous Equation to get the Homogeneous “Natural” solution:

R * it   it dt  0

1
C
Differentiating both sides of the equation and multiplying by C
di
RC * i  0
dt
The solution to this equation is of the form K H  and substituting:
 at

  
RC *  K H a  at  K H   at  0
K H   at   RC * K a 
H
 at

Simplifying,

J. N. Denenberg February 4, 2013

LTI Systems Page 20

  at  RC * a  at or a = 1/RC
The Homogeneous Solution is therefore:
t

i H t   K H  RC

i H 0 
2
But R since the voltage across the capacitor starts at 0 and,
t
2 
i H t    RC
R
But we want the voltage out which is:
t
1 2  RC
v H t  
CR
 dt
Substituting,
t t
 
v H t   *  RC )  RC  k  2 *  RC  k
2
RC where k is the constant of integration
Now we need to find the Particular Solution that is due to the Forcing Function (input)
Case 1 (t < 0): obviously, the output is again zero.
Case 2 (0 < t < 1)
We have that the original input is a constant “2” which was differentiated and became 0
The output needs to be of the form
i p t   A  B * t
Substituting into our original differential equation:
di
RC * i  0
dt
Or,
RC * B    A  B * t   0
Since this must be true for all 0 < t < 1, B = 0 (from the t term) and from the constant term
A also is 0 so
i p t   0
Therefore the total solution is
t

vt   2 *  RC
 k but this must be zero at t = 0
0

0  2 *  RC
 k  2 *1  k
k 2
 
t

vt   2 * 1   RC 

  for 0 < t < 1

J. N. Denenberg February 4, 2013

LTI Systems Page 21

Now for Case 3 (t > 1)

The Capacitor now discharges through the resistor from its value at t = 1.
 
1
  RCt
vt   2 * 1   RC  *

  And once again we get the same solution.

J. N. Denenberg February 4, 2013