Signals, Systems and

Control: 580.222
Michael Miller, Sri Sarma, Rene Vidal
January 28, 2015

Contents
I

Dynamic Systems

Linearity and Time Invariance

1.1 Examples: Constant Coefficient
Differential and Difference Equations . . . . . . . . . . . . . . .
1.2 Linearity . . . . . . . . . . . . .
1.3 Time Invariance . . . . . . . . .

4
5
7

Unit Sample and Unit Impulse Responses

of LTI Systems
10
2.1 Unit Sample Response . . . . . 12
2.2 Differential/Difference Equations,
Integrating Factors, Impulse Responses . . . . . . . . . . . . . 14
2.3 Commutatitive Property of Convolution . . . . . . . . . . . . . 19
2.4 Unit Step Response . . . . . . . 19

Properties of LTI Systems

3.1 Causality . . . . . . . . . . . .
3.2 Bounded-Input Bounded-Output
(BIBO) Stability . . . . . . . . .

22
22

Eigenfunctions of LTI Systems

4.1 The Eigenfunction property . . .

25
25

23

II Transfer Functions of LTI Systems

29
5

Fourier Transforms
5.1 The Transfer Function . . . . .
5.2 Periodic Signals & Fourier Series Expansions . . . . . . . . .
5.3 Non-periodic Fourier Transform

30
30
31
33

Properties of Fourier Transforms

6.1 Even-Odd Property . . . . . . .
6.2 The Time Shift Property . . . .
6.3 The Convolution Property and System Transfer Property . . . . . .
6.4 Transfer Function of Differentiation as an LTI System . . . . .
6.5 Remind Ourselves of What We
Know . . . . . . . . . . . . . .

37
37
37

Duality and Modulation

7.1 The Duality Property . . . . . .
7.2 The Modulation Property . . . .

44
44
45

39
41
43

Part I
Dynamic Systems

Chapter 1
Linearity and Time
Invariance
1.1

Examples: Constant Coefficient Differential and Difference Equations

Given a first order, constant coefficient differential equations with zero-initial condition
y(t)
+ ay(t) = u(t) , y() = 0 .
Let us call u(t) the input, y(t) the output, and the
system transformation equation describing the
relationship between input and output is given
by the differential equation.
Given is a first order, constant coefficient difference equation with zero-initial condition
y[n] = y[n 1] + u[n] , y[] = 0 .
Let us call u[n] the input, y[n] the output, and the
system transformation equation describing the
relationship between input and output is given
by the differential equation.
Such systems enjoy two of the most important properties we shall study through this class,
the properties of linearity and time-invariance.
(i) Linearity: We will define this subsequently,
but for now, linearity corresponds to the the very
straightforward notion that if you scale the input
u it scales the output y, and if you add inputs,
4

the outputs add as well.

(ii) Time-Invariance: This corresponds to
the notion that shifting the input in time, simply
shifts the output the same amount of time with
no other transformation.
The first order differential and difference equations have these two properties because (i) derivatives appear linearly in the equation, and (ii) the
coefficients are constant in time.

1.2

Linearity

Now let us return to the two important properties

of linearity and time-invariance. Linearity is a
very powerful property, it essentially allows for
us to divide complex signals into simpler ones
and solve more complex signals by solving the
simpler parts. Let us define linearity as follows.
Definition 1.2.1. A continuous system with input u(t) with output y(t) is linear if
u(t) creates y(t) then au(t) has output ay(t),
and
if u1 (t) creates y1 (t) and u2 (t) creates y2 (t),
then a1 u1 (t) + a2 u2 (t) creates a1 y1 (t) +
a2 y2 (t) for any u1 , u2 , a1 , a2 .
A discrete time system with input x[n] with
output y[n] is linear if
u[n] creates y[n] then au[n] has output ay[n],
and
if u1 [n] creates y1 [n] and u2 [n] creates y2 [n],
then a1 u1 [n] + a2 u2 [n] creates a1 y1 [n] +
a2 y2 [n] for any u1 , u2 , a1 , a2 .
Example 1.2.1. The first order differential equation with zero initial conditions is linear. Assume u1 creates y1 and u2 creates y2 , y1 () =
y2 () = 0, then we have to prove for input

u3 = 1 u1 + 2 u2 then output y3 = 1 y1 + 2 y2
satisfies the differential equation. Then we have
d
(1 y1 (t) + 2 y2 (t)) +a (1 y1 (t) + 2 y2 (t))
{z
}
|
{z
}
dt |
y3 (t)

y3 (t)

= 1 (y 1 (t) + ay1 (t)) + 2 (y 2 (t) + ay2 (t))

= 1 u1 (t) + 2 u2 (t) .
|
{z
}
u3 (t)

Example 1.2.2. Let the input be u(t) = eit ,

1
then y(t) = a+i
eit .
i
eit and then
To see this, y(t)
= a+i
y(t)
+ ay(t) =

i it
a
e +
eit = eit .
a + i
a + i

Example 1.2.3. So now lets use the linearity

property. Suppose we change the input from
eit to cos t. What is the output of the differential equation. Well, LTI systems allow us to solve
for this change in input by rewriting the cosines
in terms of the complex exponentials, and then
adding the solutions:
cos t =

eit + eit
,
2

implying that for u(t) = cos t then y(t) becomes

y(t) =

1 1
1 1
eit +
eit
2 a + i
2 a i

This reduces to using the identity A + A =

2Real(A) giving
y(t) =

a2

1
cos(t arctan/a) .
+ 2

Example 1.2.4. Change the input to sin t to

the differential equation using the Euler identity
it
it
sin t = e e
giving
2i


1
1
1
it
it
y(t) =
e
e
2i a + i
a i
1
a
=
sin(t arctan/a) .
a2 + 2
with (a) following from A A = 2Imag(A).
6

1.3

Time Invariance

There is a second property which this first order differential equation enjoys, the time invariance property, coming from the constant coefficients. As for the time invariance, simply stated,
an shift in time of the input results in the same
output just shifted in time.
Definition 1.3.1. A continuous time system with
input u(t) and output y(t) is time-invariant if
u(t t0 ) creates output y(t t0 ), for all inputs
u and shifts t0 . A discrete time system with input
u[n] and output y[n] is time-invariant if u[nn0 ]
creates output y[n n0 ], for all inputs u and
shifts n0 .
Example 1.3.1. Let the input be u2 (t) = ei(tt0 ) =
u1 (t t0 ), then we have to show
y2 (t) = y1 (t t0 ) =
To see this, y 2 (t) =

1
ei(tt0 ) .
a + i

i
ei(tt0 )
a+i

and then

i i(tt0 )
a
e
+
ei(tt0 )
a + i
a + i
= ei(tt0 ) = u1 (t t0 ) = u2 (t) .

y 2 (t) + ay2 (t) =

Thus we see that a simple shift of the complex exponential, the output is shifted in the same
manner. Demonstrate to yourself that the same
is true for arbitrary periodic P
inputs represented
via the Fourier series u(t) = k ak eik t .
Example 1.3.2. Let y[n] = nu[n]. Here we
can see this is not time-invariant. Pick a signal
u1 [n] = 0 for n 6= 0, and u1 [0] = 1. Then for
u1 [n] we have y1 [n] = 0. For u2 [n] = u1 [n 1]
we have y2 [1] = 1 and y2 [n 6= 1] = 0. Notice
not a simple shift so that for u2 [n] = u1 [n n0 ],
then y2 [n] 6= y1 [n n0 ].
Example 1.3.3. Let system be a constant coefficient difference equation with zero initial condition y[] = 0 and with input u[n] and output
y[n] given by system equation
u[n] = y[n] y[n 1] .
7

Lets show that it is linear, time-invariant.

Linearity: Let u1 [n] give y1 [n], and u2 [n] give
y2 [n], then we have to show for u3 [n] = au1 [n]+
bu2 [n] the solution of the difference equation is
y3 [n] = ay1 [n] + by2 [n]:
(ay1 [n] + by2 [n]) (ay1 [n 1] + by2 [n 1])
= ay1 [n] ay1 [n 1] + by2 [n] by2 [n 1] .
{z
} |
{z
}
|
au1 [n]

bu2 [n]

Time invariant: Let u1 [n] produce y1 [n], then introducing shift m0 we must prove u2 [n] = u1 [n
m0 ] produces y2 [n] = y1 [nm0 ]. Given u1 [n] =
y1 [n] y1 [n 1], then we have
u1 [n m0 ] = y1 [n m0 ] y1 [n m0 1] ;
| {z }
| {z }
|
{z
}
u2 [n]

y2 [n]

y2 [n1]

thus y2 [n] = y1 [n m0 ].
Example 1.3.4. So lets see linearity and timeinvariance in action. Let system be a constant
coefficient difference equation at rest (zero initial condition)
u[n] = y[n] y[n 1], y[] = 0 .
Then for u1 [n] = ein then the solution y1 [n] =
1
ei0 n . To see this,
1ei0
1
1
i0 n
e

ei0 (n1)
(1.1)
i
i0
0
1

e
1

e


1
i0
=
(1 e ) ei0 n . (1.2)
1 ei0

ei0 n =

Let input u[n] = cos 0 n then linearity allows us to guess the solution as additive


1
1
1
i0 n
i0 n
y[n] =
e
+
e
2 1 ei0
1 ei0


1
i0 n
= Real
e
1 ei0
1
sin 0
=p
cos(0 n tan1
).
1 cos 0
(1 cos 0 )2 + ( sin 0 )2

Summary We have examined several simple properties that the differential equation satisfies, linearity and time-invariance. Therefore,
we have arrived at the second major component
of our study of linear, time-invariant (LTI) systems, which is to understand the outputs of LTI
systems to arbitrary inputs. To do so, one needs
to understand the convolution integral and the
remaining 12 lectures work to generalize and
strengthen the these very notions.

Chapter 2
Unit Sample and
Unit Impulse
Responses of LTI
Systems
Lets examine some special signals, first in discrete time, then in continuous time.
Definition 2.0.2. The discrete time unit step is
given by
us [n] = 1 , n = 0, 1, 2, ... ;
us [n] = 0 , n < 0 .
The discrete unit sample is [n] = 1, n = 0 and
0 otherwise, defined in terms of the unit-step as
[n] = us [n] us [n 1] .
The continuous time version has similar form:
us (t) = 1 , t 0 ;
us (t) = 0 , t < 0 .
The continuous unit impulse defined in terms of
the unit-step as
(t) = u s (t)
Proposition 2.0.1 (Anti-Derivative, Anti-Difference Property). The summability or integrability

10

property is as follows:
n
X

us [n] =

[k]

k=
Z t

( )d .

us (t) =

To see this is true, take

n
n
X
X
[k] =
(us [k] us [k 1])
k=

k=

= us [n] us [n 1] + us [n 1] us [n 2] + ....
{z
}
|
telescoping series

= us [n] us [] = us [n] .
For the continuous version, we integrate the total
differential:
Z t
Z t
u s ( )d
( )d =

Z t
=
dus ( ) = us (t) us () = us (t) .

Proposition 2.0.2 (The Sifting Property). The

sifting property is true for all discrete signals
u[]:
u[n] =
u[n] =

X
k=

u[n k][k]
u[k][n k] .

k=

For the continuous time unit impulse it follows,

for smooth functions with u() = u() = 0,
Z
u(t) =
u(t )( )d

Z
u(t) =
u( )(t )d .

For the discrete version unit sample, [k] =

us [k] us [k 1] implying the sifting property:

u[n k][k] =

k=

X
k=0

u[n k]

u[n k](us [k] us [k 1])

k=

X
k=1

11

u[n k] = u[n]

For the continuous unit sample we have to use

integration by parts. Assuming the function are
differential and vanishes at , u() = u() =
0, integration by parts gives
0 = u(t )us () u(t + )us ()
Z
d(u(t )us ( ))
=

Z
Z
u(t )u s ( )d .
(du(t ))us ( ) +
=

This gives
Z

u(t )( )d =

du(t )
0

= (u(t ) u(t)) = u(t) .

2.1

Unit Sample Response

Now we define the unit sample and unit impulse

responses of our systems.
Definition 2.1.1. Define y[n] = hk [n] to be the
unit sample response of a system with input u[n] =
[n k], the unit sample shifted to time k. If
the system is time invariant, then define h0 [n] =
h[n], and hk [n] = h[n k].
Define y(t) = h (t) to be the unit impulse
response of a system with input u(t) = (t ),
the unit impulse shifted to time . If the system
is time invariant, then define h0 (t) = h(t), and
h (t) = h(t ).
Given a linear system, then the unit sample
and unit impulse responses determine the output
of these linear systems. If the systems are also
time invariant, then there is only one impulse response and it just gets shifted around.
Example 2.1.1. Return to the simple example
y[n] = nu[n]. Then h0 [n] = 0, with h1 [n] =
1[n 1], and h2 [n] = 2[n 2], ...., hk [n] =
k[n k]. We see that there are many different
impulse responses, which are not simply shifts of
each other.
12

So now we can write down the general time

solution for arbitrary inputs. It involves the convolution integral or sum of the input and the response of the system.
Definition 2.1.2. Define the convolution of two
continuous time signals f (t) g(t) as
Z
f ( )g(t )d ;
f (t) g(t) =

define their discrete analogue f [n] g[n] to be

f [n] g[n] =

f [k]g[n k] .

k=

Proposition 2.1.1. Let u[n] be rewritten as u[n] =

P
k u[k][nk]. Then the output of a linear system with unit sample responses hk [n] to input is
given by
X
y[n] =
u[k]hk [n] ;
k

if the system is time-invariant, with h0 [n] = h[n],

then
X
u[k]h[n k] .
y[n] =
|k

{z

u[n]h[n]

R
Let u(t) be rewritten as u(t) = u( )(t
)d . Then the output of a linear system with
unit sample responses h (t) to input is given by
Z
y(t) =
u( )h (t)d ;

if the system is time-invariant, with h0 (t) = h(t),

then
Z
y(t) =
u( )h(t )d .

{z
}
|
u(t)h(t)

13

2.2

Differential/Difference Equations, Integrating Factors,

Impulse Responses

Now we show that integrating factors from differential equations define impulse responses from
Signals, Systems and Controls. For particular
cases, you guys already know this!
Proposition 2.2.1. The output of this first order
differential equation at rest
u(t) = y(t)
+ ay(t), y() = 0 ,
with input u(t) is given according to
y(t) = u(t) eat us (t) ,
with unit impulse response h(t) = eat us (t).
The output of the first order difference equation at rest
u[n] = y[n] y[n 1], y[] = 0 ,
with input u[n] is given by the convolution
y[n] = u[n] n us [n] ,
with unit sample response h[n] = n us [n].
For this, examine the differential equation
and introduce the integrating factor f (t) which
has the property that it makes one side
R t of the
equation into a total differential so that dg( ) =
g(t)g(). Starting with u(t) = y(t)+ay(t),

then define the integrating factor f to satisfy

f (t)u(t) = f (t)y(t)
+ f (t)ay(t)
d
(f (t)y(t)) ,
=
dt
implying
Z t
f ( )u( )d = f (t)y(t) f ()y()

= f (t)y(t) .
14

where we use the boundary condition y() =

0. Solving the integrating factor gives
f (t)y(t)
+ af (t)y(t) = f(t)y(t) + f (t)y(t)
,
implying af (t)y(t) = f(t)y(t) .
The integrating factor becomes f = eat , allowing us to solve the differential equation according to
Z t
1
f ( )u( )d
y(t) = f (t)

Z t
ea(t ) u( )d
=
Z

u( ) ea(t ) us (t ) d
=
|
{z
}

h(t )

To derive the impulse response, by definition

set u(t) = (t), then y(t) = h(t) giving
Z t
at
h(t) = e
ea ( )d
Z
a
ea(t ) s(t )( )d
=

at

= e

us (t) ,

where in (a) we have introduced the unit step

function, and the definition which is zero to the
left of the origin, and 1 elsewhere, and we have
applied the sifting property.
For the difference equation introduce the summing factor f [n] which has the property that it
makes one side of the equation into aPtotal differ.
ence Dg[n] = g[n]g[n1] so that nk= Dg[k] =
g[n] g[]. Starting with u[n] = y[n]
y[n 1], then define the summing factor f
to satisfy
f [n]u[n] = f [n]y[n] f [n]y[n 1]
a
= f [n]y[n] f [n 1]y[n 1] ,
where (a) is the definition of the summing function. This implies
n
X

f [k]u[k] = f [n]y[n] f []y[]

k=

= f [n]y[n] .
15

where we used the boundary condition y() =

0. Solving the summing function
f [n]y[n] f [n]y[n 1]
= f [n]y[n] f [n 1]y[n 1] ,
which implies
f [n]y[n 1] = f [n 1]y[n 1] .
The summing factor becomes f [n] = n , allowing us to solve the difference equation directly according to
y[n] = f

[n]

n
X

f [k]u[k]

k=

X
k=

u[k] nk s[n k]
|
{z
}
h[nk]

with unit sample response h[n] = n us [n].

Example 2.2.1 (Adding initial condition makes
it not linear in u). Examine the same differential
equation with y(t0 ) = A. Then,
Z t
a(tt0 )
y(t) = Ae
+
u( )ea(t ) d .
t0

Taking the derivative gives

Z t
a(tt0 )
y(t)
= aAe
a
u( )ea(t ) d +u(t)
t0

which solves the differential equation y(t)+ay(t)

=
u(t) with initial condition y(t0 ) = A.
Example 2.2.2. Here is an integrating factor for
the non time-invariant case. One of the wonderful students pointed this out. Let y(t0 ) = A an
initial condition with input applied after t0 with
differential equation which has a non constant
coefficient:
u(t) = y(t)
+ a(t)y(t), y(t0 ) = A ,
then the integrating factor satisfies
f(t) = a(t)f (t) .
16

Rt

This gives f (t) = e

tion takes the form
y(t) = Ae

Rt
t0

a( )d

t0

a()d

implying the solu-

u( )e

Rt

a()d

d .

t0

To see this satisfies the differential equation

just calculate. At t0 , y(t0 ) = A, then differentiating gives
y(t)
+ a(t)y(t)
= a(t)Ae
|

Rt
t0

a( )d

Z
a(t)
{z

u( )e

Rt

a()d

t0

a(t)y(t)

+u(t) + a(t)y(t)
= u(t) .
We make the following comment. Notice
the output is a function of the input convolved
with a property of the system, eat us (t). This
property we call the impulse response of the
system and we will study it extensively. For LTI
systems this will always be true, although the
property of the system will change depending
on the system. Now we solve for the impulse
response directly.
Example 2.2.3. Lets check our answers returning to the differential and difference equations.
For
u(t) = y(t)
+ ay(t) , y() = 0 ,
with impulse response h(t) = eat us (t), lets
verify this satisfies the definition of the unit impulse response for the differential equation:

(t) = h(t)
+ ah(t) , h() = 0 .
For this we will use the fact that u s (t) = (t).
Assuming h(t) = eat us (t), we have

h(t)
+ ah(t)
d at
e us (t) + aeat us (t)
=
dt
d
= eat us (t) aeat us (t) + aeat us (t)
dt
d
=
us (t) = (t) .
dt
17

d
}

Example 2.2.4. Lets check the unit sample response for the difference equation
u[n] = y[n] y[n 1] , y[] = 0 ,
then lets see
h[n] = n us [n] .
satisfies definition of unit sample response
[n] = h[n] h[n 1] .
Then h[n] = 0, n 1, then
h[1]
h[0]
h[1]
h[2]
h[n]

=
=
=
=
=

0
[0] = 1
h[0] =
h[1] = 2
n , n 0 .

Example 2.2.5. Examine an R-C circuit with input voltage Vinput (t) with the resistor and capacitor in series and taking the output across the
capacitor. Then
Vinput (t) = Vr (t) + Vc (t) ;
defining I(t) the current through the resistorcapcitor then the voltage across the resistor Vr (t) =
RI(t) and the voltage across the capactitor using the relationship Q = CV gives
Z
1 t
I( )d .
Vc (t) =
C
Then taking the voltage across the capacitor
as the output y(t) with the input voltage source
u(t). Taking the derivative of the output gives
an expression for the current, I(t) = C y(t).

Adding voltages in terms of u(t) input and y(t)

output gives the differential equation
u(t) = RC y(t)
+ y(t) , y() = 0 .
Notice the initial condition corresponds to no
voltage across the capacitor - system at rest.
18

We have our LTI system (first-order constantcoefficient differential equation), with impulse
t
1 RC
response h(t) = RC
e
us (t) (notice, the 1/RC
scaling results from the equation). The output
becomes
Z
1 t
u( )
e RC us (t )d
y(t) =
RC

Z t
t
1
=
u( )e RC d .
RC

2.3

Commutatitive Property of
Convolution

Proposition 2.3.1. Convolution has the property

that
u(t) h(t) = h(t) u(t) ;
u[n] h[n] = h[n] u[n] .
Show this by redefining the limits of the integral. Lets do the difference equation.

u[k]h[n k]

nk=r

k=

X
r=

u[n r]h[r]
h[r]u[n r]

r=

2.4

h[n] u[n] .

Unit Step Response

We can derive the impulse response from the

unit step response by using the fact that derivative and differences of step response gives the
impulse response.
Proposition 2.4.1. Given a continuous time LTI
system with input u(t) = us (t) the unit step,
then the derivative of the response is the impulse
response:
y(t)
= h(t) .

19

Given a discrete time LTI system with input

u[n] = us [n] the unit step, then the first difference of the response is the impulse response:
y[n] y[n = 1] = h[n] .
Proof.
d
d
us (t) h(t) = h(t) us (t)
dt Z
dt
d
=
h( )us (t )d
dt
Z
d
=
h( ) us (t )d
dt
Z
=
h( )(t )d = h(t) .

y(t)

y[n] y[n 1]
= (us h)[n] (us h)[n 1] = (h us )[n] (h us )[n 1]
X
X
h[k]us [n 1 k]
h[k]us [n k]
=
k

h[k](us [n k] us [n 1 k])

h[k][n k] = h[n] .

Example 2.4.1. So now we have our general solutions for the difference equations and differential equations, so we can look at unit-step responses. For the difference equation
u[n] = y[n] y[n 1] ,
we have
X

y[n] = u[n] h[n] =

=
=

k=

X
k=
n
X
k=

20

u[k]h[n k]
u[k]nk us [n k]
u[k]nk .

For an input of u[n] = us [n], then we have

y[n] =

n
X

us [k]nk =

k=

n
X

nk us [n]

k=0
(n+1)

1
us [n]
1 1
1 n+1
n+1 1
us [n] =
us [n]
=
1
1
= n

For the differential equation

u(t) = y(t)
+ ay(t)
we have
Z

u( )ea(t ) us (t )d
y(t) = u(t) h(t) =

Z t
u( )ea(t ) d .
=

For u(t) = us (t), then

Z t
1
y(t) =
ea(t ) d us (t) = eat ea |t0 us (t)
a
0
1
=
(1 eat )us (t)
a
Example 2.4.2. Lets check the unit-step response
result works for the impulse response of an LTI
system using the difference formula h[n] = y[n]
n+1
us [n]. Then
y[n 1] with y[n] = 1
1
1 n+1
1 n
us [n]
us [n 1]
1
1
1 n+1 1 + n
=
1
n (1 )
=
= n
1
1
=
=1.
1

y[n] y[n 1] =
n>0

n=0

So h[n] = y[n] y[n 1] = n us [n].

21

Chapter 3
Properties of LTI
Systems
3.1

Causality

Definition 3.1.1. A continuous time system is

causal, if the output y(t) at time t is not a function of future inputs. A discrete time system is
causal if the output y[n] at time n is not a function of future inputs.
Here is a necessary condition for LTI systems.
Proposition 3.1.1. If an LTI system is causal,
then this implies h(t) = 0, t < 0. For discrete
systems, h[n] = 0, n < 0.
To see this,
Z

u( )h(t )d

y(t) =

u( )h(t )d

+
t

= f (u( ), > t) ,
if h(t) 6= 0 f or t < 0 .
Example 3.1.1. Let h(t) = us (t + 3), then system is not causal.
Z
y(t) = us (t) h(t) =
us ( )us (t + 3 )d

Z t+3
=
us ( )d

= f (us (s), s > t) .

22

3.2

Bounded-Input BoundedOutput (BIBO) Stability

Definition 3.2.1. A system is BIBO stable (boundedinput, bounded output), if a bounded input implies a bounded output.
Here is a sufficient condition for LTI systems.
R
Proposition 3.2.1. P
If |h(t)|dt < , or for
discrete systems if
n= |h[n]| < then the
system is BIBO stable.
Let |u(t)| < B, t, then
y(t) = u(t) h(t) = h(t) u(t)
Z
=
h( )u(t )d

Z
|h( )||u(t )|d

Z
B
|h( )|d < C

For all t,|y(t)| C.

Example 3.2.1. Let h(t) = us (t), then the system is not BIBO stable. Let input u(t) = us (t),
then |us (t)| 1 bounded, and
Z
us ( )us (t )
y(t) = us (t) us (t) =

Z t
d = t .
=
0

There does not exist a bound C for all time t.

Example 3.2.2. Take our favorite differential equation, with 0. This doesn not satisfy the
sufficient condition for stability. Then h(t) =
et us (t) and for 0 we have
Z
1
et dt = (e 1) = .

0
Notice for = 0, you have to be more careful (differentiate numerator and denominator).
23

Example 3.2.3. Take the difference equation u[n] =

y[n] y[n 1], y[] = 0; then for unit step
input u[n] = us [n],
y[n] =

1 n+1
us [n] .
1

For interest rate = 1.07 then notice this is not

stable. Bounded input u[n] = us [n] gives output
that has no bound growing at rate 1.07n .

24

Chapter 4
Eigenfunctions of
LTI Systems
So now we have examined several simple properties that systems satisfy including linearity and
time-invariance, and the implication that the outputs for LTI systems are given by convolution
integrals and sums. Now we will see that complex exponentials have the special property that
they pass through changed only by a complex
number in the differential equation. We will see
the roll of transforms, as representing arbitrary
inputs via the superpositions of complex exponentials. This discussion is often called a frequency domain analysis. Frequency domain
analysis studies the outputs of linear and timeinvariant systems via their response to complex
exponentials.

4.1

The Eigenfunction property

The differential equation is a beautiful example of a linear and time-invariant system, denoted LTI. All LTI system enjoy the following
exquisite property which lays the basis for our
use of Transforms and the special roll of complex exponentials, cosines and sines in LTI system.
Example 4.1.1. Let us see that this is true for
our differential equation. Let u(t) = ei2f t , then
the output to the differential equation with boundary condition y() = 0 is simply a scaled
25

version of the input:

y(t) = H(f )ei2f t = H(f )u(t) .
The complex coefficient H(f ) is independent
1
.
of time t and given by H(f ) = a+i2f
To see this, simply plug it into the differential
equation, giving for u(t) = ei2f t then
t
e|i2f
H(f )ei2f t + aH(f )ei2f t ,
{z } = i2f
|
{z
} |
{z
}
u(t)

y(t)

which implies H(f ) =

ay(t)

1
.
a+i2f

Example 4.1.2. Not all inputs have this property. Let us look at a particular example, examine the special function, the sinc function,
very important in communcations engineering
sin 2t
.
2t
Let us demonstrate that y(t) 6= Hu(t) for any
complex constant H. Assume it is, then solving
the differential equation gives
u(t) = sinc(t) =

cos 2t
sin 2t
sin 2t
sin 2t
H
+aH
=
.
2
2t
2t
2t
2t
Choosing t = 21 implies H = 0, which is a contradiction since for other times t 6= 0 the value
H = 0 doesnt solve the equation.
H2

So, the fact that complex exponentials pass

through this differential equation unchanged is
in fact a general property of all LTI systems, and
is an important result. It forms the basis for the
application of Fourier and Laplace transforms to
LTI systems.
The general result is that LTI systems transform complex exponentials with a simple complex number; this complex number turns out to
be the Fourier transform of the system impulse
response.
Proposition 4.1.1. For input u(t) = ei2f t to
the LTI system with impulse response h(t) and
output y(t), then
Z
i2f t
y(t) = e
h( )ei2f d ;
{z
}
|
Fourier transform

26

for input u[n] = ei2f n and discrete LTI system

then output
y[n] = e

i2f n

h[k]ei2f k

k=

{z

discrete Fourier transform

We will call these complex numbers the Fourier

tranform H of the impulse response h.
Proof. To prove it, simply do the continuous time:
y(t) = u(t) h(t) = h(t) u(t)
Z
=
h( )u(t )d

Z
h( )ei2f (t ) d
=

Z
i2f t
= e
h( )ei2f d .

Example 4.1.3. Given the differential equation

u(t) = y(t)
+ ay(t) ,
then h(t) = eat s(t) giving
Z
Z
i2f
h( )e
d =

e(a+i2f ) d

e(a+i2f ) |
0
.
a + i2f

For a > 0, then the upper limit is zero and the

result becomes
Z
1
H(f ) =
e(a+i2f ) d =
.
a + i2f
0
You have now computed your first Fourier
transform.
Example 4.1.4. Given the difference equation
with || < 1, then
u[n] = y[n] y[n 1] ,

27

then
H(f ) =
=

h(k)ei2f k

k=

(ei2f )k =

1
.
(1 ei2f )

CHECKING OUR ANSWERS: Notice, complex exponentials pass right through this system
as well with just a complex number added. guess
y[n] = H(f )ei2f n , then
ei2f n = H(f )ei2f n H(f )ei2f (n1) ,
implying
H(f ) =

1
1 ei2f

28

Part II
Transfer Functions of
LTI Systems

29

Chapter 5
Fourier Transforms
5.1

The Transfer Function

We have already seen the special property of

complex exponentials: they pass through relatively unchanged LTI systems, certainly the differential equation. Laplace and Fourier transforms exploit this idea, realizing that there are
many signals that can be represented via complex exponentials. So, let us start with the Fourier
series, a simple idea: suppose arbitrary signals
could be represented according to the complex
exponential series given by
X
u(t) =
Uk ei2fk t , Uk complex .
k

Notice, if T is the period of each of the frequencies so that fk = k/T (which will be the
case when we look at Fourier series), then the
signal x(t) in this expansion is periodic u(t) =
u(t mT ).
The special property of LTI systems is that
using the divide and conquer strategy of LTI
systems, the output of such a series is a linear
superposition of complex sinusoidal inputs.
Example 5.1.1. For y(t) related to input u(t)
with differential equation at rest y(t)
P+ ay(t) =
u(t), y() = 0, then for u(t) = k Uk ei2fk t
the output becomes
X
1
ei2fk t .
y(t) =
Uk
a
+
i2f
k
| {z k}
Fourier transform H(f)

30

For an LTI system, u(t) = ei2f t , then the

complex number determining the output y(t) =
H(f )ei2f t is given by the Fourier transform of
the impulse response, also called the transfer function of the system;
Z
h(t)ei2f t dt .
H(f ) =

The transfer function formula for arbitrary

inputs is given as follows.
Proposition 5.1.1. Given inputs
X
u(t) =
Uk ei2fk t ,
k

to LTI system with impulse response h(t), then

notice, outputs of LTI systems y(t) always take
the form:
X
y(t) =
Uk H(fk ) ei2fk t ,
| {z }
k

system transfer

where
Z

h(t)ei2fk t dt .

H(fk ) =

This is why transform representations are so

valuable.

5.2

Periodic Signals & Fourier

Series Expansions

Many, many signals have complex sinusoidal representations.

First we notice that series representations built
from harmonics of frequencies f0 = T1 are periodic on T ! Let us now throughout denote periodic signals with period T via the notation uT (t),
explicitly indicating the period.
Proposition 5.2.1. Let uT (t) be a superposition
of harmonics fk = Tk = kf0 , so that
X
X
k
uT (t) =
Uk ei2 T t =
Uk ei2kf0 t ,
k
T

k
T

then, u (t) = u (t mT ).
31

The Fourier series representations gives a formula for the complex numbers in the representations Uk .
Proposition 5.2.2. Let uT (t) be periodic with
period T , so that the frequencies fk = Tk =
kf0 ,f0 = 1/T and
SYNTHESIS
X
X
k
u (t) =
Uk ei2 T t =
Uk ei2kf0 t
T

Then, the complex coefficients Uk are given

by

Uk

ANALYSIS
Z
k
1 T T
u (t)ei2 T t dt
=
T 0
Z
1 T /2 T
u (t)ei2kf0 t dt .
=
T T /2

We denote these as a Fourier series pair:

uT (t) Uk .
Proof: Let prove the analysis equation for
the complex coefficients. Use the property that
R T i2 (mn) t
T
e
dt = T [m n]. Then we have
0
Z

uT (t)ei2 T t dt

=
0

X
k

Uk ei2 T t ei2 T t dt

k
T

ei2

Uk

(km)
t
T

dt

Uk T [k m]

OK, so how do we use this. Well, for periodic signals with period T , then we just have to
evaluate the Fourier series coefficients Uk .
Example 5.2.1.
1. u(t) = constant, then a0 =
constant and Uk = 0, k 6= 0 for any period T .
32

2. uT (t) = ei2f0 t with T =

Uk = 0, k 6= 1.

1
,
f0

3. uT (t) = cos 2f0 t, then T =

a1 = 21 , Uk = 0, k 6= 1.
4. uT (t) = sin 2f0 t, then T =
a1 = 2i1 , Uk = 0, k 6= 1.

5.3

1
,
f0

a1 = 1,
1
,
f0

a1 =

a1 =

1
,
2i

Non-periodic Fourier Transform

So, Fourier series is for periodic signals; the Fourier

transform is for non-periodic signals. Lets examine and construct the Fourier transform by allowing the period of the periodic signals to go to
, see what we get.
Lets define uT (t) to be the periodic version
of u(t), where u(t) has finite support u(t) =
0, |t| T /2. Thus,
u(t) = uT (t), t [T /2, T /2]
u(t) = 0 , otherwise
Definition 5.3.1. Define the Fourier transform
of u(t) to be
Z
u(t)ei2f t dt .
U (f ) =

Proposition 5.3.1. Then for signals u(t) which

are zero outside of T /2, T /2, then we have the
relationship between FT and FS.
1
1
Uk =
U (f )|f =kf0 f0 =
T
XT
T
where u (t) =
Uk ei2kf0 t .
k

Proof.
Uk

Z
k
1 T /2 T
u (t)ei2 T t dt
=
T T /2
Z
k
1
=
u(t)ei2 T t dt
T
1
=
U (f )|f = k .
T
T

33

Example 5.3.1. Let uT (t) = B, t [ A2 +

mT, A2 + mT ] and 0 otherwise, with m an integer. Then
Uk

T /2

1
=
T

B
=
T

uT (t)ei2 T t dt

T /2
A
2

ei2 T t dt

A
2
k

i2 T t 2
| A
Be
2

i2 Tk

B sin Tk A
B ei T A ei T A
=
=
T
k
i2 Tk
Example 5.3.2. Let u(t) = B, t [ A2 , A2 ], and
0 otherwise. Then
Z
u(t)ei2f t dt
U (f ) =

A
2

= B

ei2f t dt

A
2
A

= a

ei2f t |2 A
2

i2f
B sin f A
=
f

=B

eif A eif A
i2f

Let uT (t mT ) = u(t), t [T /2, T /2].

Then
X
1
uT (t) =
Uk ei2kf0 t where f0 =
T
k
1
B sin kf0 A
U (kf0 ) =
T
T kf0
k
B sin T A
1
=
= U (f )|f = k
T
k
T

Uk =

OK, so we see that the Fourier transform can

be used to define the Fourier series. Now what
we would like to do is understand how to represent the periodic signals when the period goes
to infinity T , so that the we can have a
synthesis pair.

34

Proposition 5.3.2. Then the Fourier transform

is given by
ANALYSIS
Z
u(t)ei2f t dt ,
U (f ) =

with synthesis equation

SYNTHESIS
Z
U (f )ei2f t df .
u(t) =

Proof: Lets remind ourselves that uT (t) is

the periodic version of u(t), where u(t) has finite support u(t) = 0, |t| T2 . Let uT (t) be periodic with period T , and u(t) = limT uT (t).
Then
X
u(t) = lim uT (t) = lim
Uk ei2kf0 t
T

X1
U (kf0 )ei2kf0 t
= lim
T
T
k
X
= lim
U (kf0 )ei2kf0 t f0
f0 0

U (f )ei2f t df

Example 5.3.3. Let u(t) = B, t [ A2 , A2 ], and

0 otherwise, then the Fourier transform
Z
U (f ) =
u(t)ei2f t dt

Z A
B sin f A
=
Bei2f t dt =
.
f
A
Then the synthesis shows
Z
B sin f A i2f t
u(t) =
e
df .
f

Example 5.3.4. Another example, go back to the

impulse response of the differential equations
h(t) = eat us (t) H(f ) =

35

1
.
a + i2f

To see this the Fourier transform

Z
h(t)ei2f t dt
H(f ) =
Z

=
eat us (t)ei2f t dt =

1
a + i2f

The synthesis becomes

Z
H(f )e+i2f t df
h(t) =
Z

1
=
ei2f t df
a + i2f
Example 5.3.5. Let u(t) = (t), then
Z
U (f ) =
(t)ei2f t dt = 1

36

Chapter 6
Properties of Fourier
Transforms
6.1

Even-Odd Property

Lets remind ourselves of some properties.

Z
U (f ) =
u(t) (cos 2f t i sin 2f t)dt ,
then if u(t) = real, then RealU (f ) is even, then
ImaginaryU (f ) is odd.
Let u(t) = evensignal, so that u(t) = u(t),
then
Z
U (f ) = u(t) (cos 2f t i sin 2f t)dt
Z
Z
=
u(t) cos 2f t dt i u(t) sin 2f t dt
|
{z
}
{z
}
|
even

odd

Thus if u(t) is real and u(t) is even, then U (f )

is real.
Example 6.1.1.
A A
B sin f A
u(t) = B, t [ , ] U (f ) =
2 2
f
| {z }
Real,Even

u(t) = (t) 1

6.2

The Time Shift Property

The time shift property is important. It says that

the magnitude of the Fourier transform doesnt
37

change - just the relative phases of each of the

components. For pure time shift this corresponds
to linear phase.
Proposition 6.2.1. If u(t) U (f ), then
y(t) = u(t t0 ) Y (f ) = U (f )ei2f t0
Proof.
Z
y(t) =
Z
=
Z
=

Y (f )ei2f t df = u(t t0 )
U (f )ei2f (tt0 ) df
U (f )ei2f t0 ei2f t df
{z
}
|
Y (f )

Example 6.2.1. (t t0 ) ei2f t0 To see

this,
Z
U (f ) = (t t0 )ei2f t dt .
Example 6.2.2. ei2f0 t (f f0 ) To see
this
Z
u(t) =
U (f )ei2f t df
Z
=
(f f0 )ei2f t df = ei2f0 t
Example 6.2.3. Let uT (t) a periodic waveform
/ [ A2 , A2 ]
uT (t) = B, t [ A2 , A2 ], uT (t) = 0, t
periodic uT (t mT ) = uT (t). Then the Fourier
series uT (t) Uk in Fourier series is given
by Uk = T1
Then
uT (t) =

B sin Tk A
Tk

B sin Tk A
.
k

X B sin k A
T

ei2 T t ei2

kA/2
T


k
BA X B sin Tk A  i2 k t
=
+
e T + ei2 T t .
T
k
k=1

BA X B sin Tk A
k
=
+
2
cos(2 t) .
T
k
T
k=1
38

6.3

The Convolution Property

and System Transfer Property

A celebrated and important property, showing

the role of Fourier transforms and LTI systems.
For uT (t) = uT (tmT ) periodic so then y T (t) =
uT (t) h(t). Thus for periodic inputs, we use
Fourier series for the outputs.
Proposition 6.3.1. Convolution Property for
Fourier Series Let uT (t) be periodic, period T ,
f0 = T1
X
uT (t) =
Uk ei2kf0 t
k

then the output of an LTI system y T (t) =

with impulse response h(t) is given by

Yk ei2kf0 t

y T (t) = uT (t) h(t)

X
=
U H(kf ) ei2kf0 t
| k {z 0}
k
Yk
Z
H(kf0 ) =
h(t)ei2kf0 t dt
Proposition 6.3.2. Convolution Property for
Fourier Transform Let period go to infinity u(t) =
limT uT (t), and
Z
U (f )H(f ) ei2f t df .
y(t) = u(t) h(t) =
|
{z }

Y (f )

y(t) =
=

lim uT (t) h(t)

X
lim
Uk H(kf0 )ei2kf0 t

T
T

X1
= lim
U (kf0 )H(kf0 )ei2kf0 t
T
T
k
X
= lim
U (kf0 )H(kf0 )ei2kf0 t f0
f0 0

U (f )H(f )ei2f t df

39

Proposition 6.3.3.
u(t) U (f ) , h(t) H(f ) ,
then
y(t) = u(t) h(t) Y (f ) = U (f )H(f ) .
Proof.
Z
y(t) = u(t) h(t) = u(s)h(t s)ds
Z Z
Y (f ) =
u(s)h(t s)ds ei2f t dt
|
{z
}
y(t)(r=ts)

Z Z
=

u(s)h(r)ei2f (r+s) dsdr

= U (f )H(f )

Example 6.3.1. Examine the RLC circuit, output across the capacitor at rest, y() = 0,
then
y(t)
=

1
1
y(t) +
u(t) ,
RC
RC
t

1 RC
with impulse response h(t) = RC
e
s(t). Then
T
assume input u (t) a periodic waveform uT (t) =
B, t [ A2 , A2 ], periodic uT (t mT ) = uT (t).
Then the output y T (t) in Fourier series is given
by

y T (t) =
X 1 B sin f A
k

|f = k
T

k
1
|f = k ei2 T t .
T
1 + i2f RC

Notice, as T , then y T (t) y(t) with

Z
B sin f A
1
y(t) = df
ei2f t
f
1 + i2f RC
Example 6.3.2. LTI system y(t)

= ay(t) +
at
u(t), then h(t) = e s(t) and let input u(t) =

40

B, t [ A2 , A2 ]. Then
Z
U (f )H(f )ei2f t df
y(t) =
Z

B sin f A
1
ei2f t df
=
f
a + i2f

{z
}
|
Y (f )

= u(t) h(t)
The input is real to the differential equation
so we should expect the output will be real as
well.
Z
B sin f A
1
y(t) =
ei2f t df
f
a + i2f

|
{z
}
Y (f )

=
Z 00
+

1
B sin f A
ei2f t df
f
a + i2f
B sin f A
1
ei2f t df .
f
a + i2f

Take the second term and substitue r = f , giving the term

Z 0
1
B sin (r)A
ei2(r)t (dr)
(r)
a

i2r
+
Z
B sin rA
1
=
ei2rt dr
r
a

i2r
0
Noticing this is the complex conjugate gives the
output in terms of a real cosine series using the
formula A + A = 2Real(A) for A complex:


Z
B sin f A
1
i2f t
Real
e
df
y(t) =
f
a i2f
0
Z
2B sin f A
1
p
=
2
f
a + (2f )2
0
2f
)df
cos(2f t tan1
a

6.4

Transfer Function of Differentiation as an LTI System

Proposition 6.4.1. Differentiation y(t) = dtd u(t)

is a linear, time-invariant transformation; the
41

Fourier transform pair satisfies

y(t) =

d
u(t) i2f U (f ) .
dt

Viewed as an LTI system, the transfer function is

H(f ) =

1
Y (f )
=
.
U (f )
i2f

To see this use the defintiion

Z
d
d
y(t) = u(t) =
U (f )ei2f t df .
dt
dt
Example 6.4.1. Given our favorite differential
equation at rest y() = 0 with y(t)+ay(t)

=
u(t), then we can use the linearity of the Fourier
transform to calculate the transfer function.
y(t)
i2f Y (f ) , y(t) Y (f ), u(t) U (f ) .
Then we have the transfer function equation
i2f Y (f ) + aY (f ) = U (f ) ,
implying H(f ) =

Y (f )
U (f )

1
.
i2f +a

Example 6.4.2. Let the input be u(t) = u(t +

f A
A/2) u(t A/2), then U (f ) = sinf
and the
output of the differential equation becomes
Z
1
sin f A
ei2f df
y(t) = u(t)h(t) =
f a + i2f
|
{z
}
U (f )H(f )

42

6.5

Remind Ourselves of What

We Know

Lets collect some results.

A A
uT (t) = B, t [ , ], 0otherwise,
2 2
B sin f A

f
Z
B sin f A i2f t
u(t) =
e
df
f
A A
uT (t) = B, t [ , ], uT (t) = uT (t + mT )
2 2
X 1 B sin f A
uT (t) =
|f =kf0 ei2kf0 t
T
f
k
1
u(t) = eat s(t)
a + i2f
Z
1
eat s(t) =
ei2f t df
a + i2f
y(t)

= ay(t) + ei2f0 t at rest then

1
ei2f0 t
y(t) =
a + i2f0
Z
(t) 1 (t) =
u(t t0 )
u(t t0 )

u(t A/2)

u(t A/2)

1ei2f t df

U (f )e2f t0
Z
U (f )e2f t0 ei2f t df
B sin f A i2f A
2
e
B, t [0, A]
f
Z
B sin f A i2f A i2f t
2 e
e
df
f

Periodic with period T:

y T (t) = uT (t) h(t)

Uk H(kf0 ) f0 =

k
T

y (t) = u (t) h(t) =

Uk H(kf0 )ei2kf0 t

y(t) = u(t) h(t) U (f )H(f )

Z
y(t) = u(t) h(t) = U (f )H(f )ei2f t df

43

1
T

Chapter 7
Duality and
Modulation
7.1

The Duality Property

Let us try to understand the following relationships. Suppose we know that

B sin f A
A A
u(t) = B, t [ , ]
,
2 2
f
then what is

B sin tA
t

Proposition 7.1.1.
u(t) U (f ) then
U (t) u(f ) .
Proof of : U (t) u(f )
Beginning with u(t) U (f ) implies
Z
Z
i2f t
u(t) = U (f )e
df = U (r)ei2rt dr.
Now substituting f for the variable t gives
Z
u(f ) = U (r)ei2rf dr ,
and changing the dummy variable r to t gives
the Fourier transform pair U (t) u(f ) according to
Z
u(f ) = U (t)ei2tf dt .
44

Example 7.1.1.
A A
B sin f A
, ]
2 2
f
A A
B sin tA
u(f ) = B, f [
, ]
t
2 2
e.g.
u(t)

B, t [

u(t) = B, t [0, A]
B sin tA
t

B sin f A i2f A
2
e
f

ei2t 2 u(f ) = B, f [0, A]

Example 7.1.2.
u(t)
1
a + i2t
(t)
1
(t + t0 )
e+i2tf0
(t t0 )
ei2tf0

7.2

eat s(t)

1
a + i2f

u(f ) = eaf s(f )

1
(f )
e+i2f t0
((f + f0 )) = (f f0 )
ei2f t0
((f f0 )) = (f + f0 )

The Modulation Property

Now we have already seen convolution results in

multiplication in the frequency representations.
How about multiplication in time?
Proposition 7.2.1.
u(t) U (f )
u(t) e
U (f f0 )
i2f0 t

Proof:
Z
Z
i2f0 t i2f t
u(t)e
e
dt =
u(t)ei2(f f0 )t
= U (f f0 ) .
Proposition 7.2.2.
u(t) U (f )
,
y(t) Y (f )
then z(t) = u(t) y(t) Z(f ) = U (f ) Y (f )
45

Proof:
Z Z
z(t) =

U (r)Y (f r)drei2f t df

Z Z

U (r)Y (s)drei2(r+s)t ds
Z
Z
i2r
=
U (r)e dr Y (s)ei2s ds = u(t) y(t) .

46