EE21201/EE21101: Signals and Systems

Instructor: Prof. Ashish R. Hota

Class Hours: E4 Slot. Wednesday: 12pm - 12:55pm, Thursday: 11am -

11:55am, Friday: 9am - 10:55am
Venue: NR 322
Grading Scheme: 50 % Endsem, 30 % Midsem, 20 % Class Tests, etc.
Preferred Mode of Contact: Send email to [email protected] with
subject containing [EE21201]. Do not forget to write your name and roll
no. Any email with a blank subject and without name and roll no. will be
ignored.
Class handouts and other materials will be uploaded on MS Teams. To join
the team, use code: oettxk9
Reference Books:
1. Signals and Systems 2nd Edition, by Oppenheim, Willsky and Nawab
2. Principles of Linear Systems and Signals, by B. P. Lathi
General expectation: spend 3-4 hours every week to keep up with the lecture,
spend time solving problems yourself rather than watch videos online
Office hours: 5:15pm - 6:00pm every Wednesday

1
 Systems

System: any mapping of (subclass of) signals to signals

x(t) y(t)
System

Dynamical system: any system with an internal state and a rule govering
how the state evolves in time

Example: a car with input being acceleration a(t) and output being velocity
v(t). The evolution of speed is given by
dv
= a(t).
dt
The output is measured by the speedometer sensor. This is an example of
a continuous-time dynamical system.

The field of signal processing studies how to extract desirable features from
given signals, often via design of filters.

The field of control systems studies how to design systems (controllers)

which produce suitable signals to steer the output of a given dynamical
system in a desirable manner.

2
 Continuous Time Systems: Examples

x(t) y(t)
System

Electrical Circuits:
VS = 40V
R6

R7
C3 R4 Q2

R8

C1 C2 R2 C5
R1
Q1 Q3

output y(t)

input x(t) R3
R5 C4

Flight Control: Multiple Input Multiple Output (MIMO)

3
 Discrete Time Systems: Examples

x[n] y[n]
System

Economy:
Input at day n = federal reserves mortgage rate
Output at day n = inflation rate at day n

Sound Processing:
Input x[n] = input sound
Output y[n] = output sound

4
 Important Mathematical Systems

x(t) y(t)
System

Delay: x(t) ! y(t) = x(t t0 ) or x[n] ! y[n] = x[n n0 ]

Scaling: x(t) ! y(t) = ax(t) or x[n] ! y[n] = ax[n]
Di↵erentiator: x(t) ! y(t) = dtd x(t) or x[n] ! y[n] = x[n] x[n 1]
Rt P
Integrator: x(t) ! y(t) = 1 x(⌧ )d⌧ or x[n] ! y[n] = nk= 1 x[k]
What is the output of the DT integrator system to the input x[n] = [n]?

5
 Properties of Systems: Memoryless

x(t) y(t)
System

We say that a system is memoryless if output at time t (or n) only depends

on the value of the input at time t (or n)
In words: the mapping is instantaneous
Example 1: Is y(t) = x2 (t) memoryless?
Example 2: Is y[n] = x[n 1] memoryless?

6
 Properties of Systems: Causality

x(t) y(t)
System

A system is causal if output at time t (or n) only depends on the input at

time s  t (or k  n)
Example 1: Is the delay system y(t) = x(t t0 ) causal?
Rt
Example 2: Is the integrator system y(t) = 1 x(⌧ )d⌧ causal?
Example 3: Is y[n] = x[ n] causal?

7
 Properties of Systems: Invertible

x(t) y(t)
System

A system is invertible if distinct inputs produce distinct outputs

Equivalently: A system is invertible if from the output one knows the input
– To show that a system is invertible: construct the inversion
– To show that a system is not invertible: show two signals map to the
same output
Example 1: Is y(t) = x(2t) invertible?
Example 2: Is the system y[n] = x[n] + y[n 1] invertible?
Example 3: Is y(t) = x2 (t) invertible?

8
 Properties of Systems: BIBO Stable

x(t) y(t)
System
x[n] y[n]

Definition: We say that a signal is bounded by B > 0 if |x(t)|  B (or

|x[n]|  B) for all t (or n)
Example: Is x(t) = tu(t) bounded? What about x[n] = ejn ?
BIBO Stability: We say that a system is Bounded-Input Bounded-Output
(BIBO) stable if all the bounded inputs result in bounded outputs.
Example 1: Is y(t) = 2x2 (t 1) + x(2t) bounded?
Pn
Example 2: What about the integrator system y[n] = k= 1 x[k]?

9
 Properties of Systems: Time-Invarying (TI)

x(t) y(t) x(t t0 ) y(t t0 )

System System
x[n] y[n] x[n n0 ] y[n n0 ]

We say that a system is TI if for all inputs x(t) (or x[n]) and all time-shifts
t0 (or n0 ) the output to x(t t0 ) (or x[n n0 ]) is y(t t0 ) (or y[n n0 ])
Example 1: Is y(t) = sin(x(t)) Time-Invarying?
Example 2: Is the system y[n] = nx[n] TI?
Clicker Question: Is the system y[n] = x[n]x[n 3] TI?
a. Yes
b. No

10
 Properties of Systems: Linear

x(t) y(t)
System
x[n] y[n]

We say that a system is linear if for all x1 (t), x2 (t) (or x1 [n] and x2 [n]) and
all scalars a1 , a2 the response (output) of the system to a1 x1 (t) + a2 x2 (t) is
a1 y1 (t) + a2 y2 (t), where y1 is the response to x1 and y2 is the response to
x1
Rt
Example 1: Is the integrator system x(t) = 1 x(⌧ )d⌧ linear?
Example 2: What about y[n] = (x[2n])2 ?
Clicker Question: Is the system y(t) = (t 1)x(t + 2) linear?
a. Yes
b. No

Rt
Example: Determine the properties of the system y(t) = 1 ⌧ x(⌧ ).

11
 Discrete-time LTI Systems

Definition: We refer to the response (output) of the system to the input

x[n] = [n] as impulse response of the system.
Denote the impulse response by h[n].
Given any signal x[n]

x[n]

1 1 2

we can write it as: x[n] = . . . + x[ 1] [n + 1] + x[0] [n] + x[1] [n 1] +

x[2] [n 2] + . . .
P
In short: x[n] = 1 k= 1 x[k] [n k] (representation property)

12
 Discrete-time LTI Systems

By Time-Invariant property:

[n + 1] h[n + 1]
LTI System
[n] h[n]
LTI System
[n 1] h[n 1]
LTI System
..
.
[n k] h[n k]
LTI System

By Linearity:
P1 P1
x[n] = k= 1 x[k] [n k] y[n] = k= 1 x[k]h[n k]
LTI System

Let h[n] be the impulse response of an LTI system. Then for any input x[n],
the output is
1
X
y[n] = x[k]h[n k].
k= 1

The above sum is called the convolution (sum) of x[n] and h[n]. Denote
def
it by y[n] = x[n] ⇤ h[n]

13
 Discrete-time LTI Systems

Clicker Question Suppose that we have a DT LTI system and we know

that the response (output) to x[n] = [n] is y[n] = u[n].
1
h[n]
0.8
0.6
0.4
0.2
n
2 2 4

What is the response of the system to x[n] = u[n] u[n 3] (plotted

above)?
a. y[n] = h[n]
b. y[n] = sin( 2⇡
3 n)
c. y[n] = u[n] + u[n 1] + u[n 2]
d. none of the above

14
 Continuous-time LTI Systems

⇢ 1
if t 2 [0, ]
Reminder: Define (t) =
0 else.
1

0
0

(t) is a pulse of hight 1 and width

P1
We can approximate a nice-enough signal x(t) ⇡ x̂(t) = k= 1 x(k ) (t
k )
2
1
t
2 2
1
2

Let h (t) be the response to the input (t)

By Time-Invariant property:

(t k ) h (t k )
LTI System

By Linearity:
P1 P1
x̂(t) = k= 1 x(k ) (t k ) ŷ(t) = k= 1 x(k ) h (t k )
LTI System

15
 Continuous-time LTI Systems

P1
x(t) ⇡ x̂(t) = k= 1 x(k ) (t k )
2
1
t
2 2
1
2

By LTI property:
P1 P1
x̂(t) = k= 1 x(k ) (t k ) ŷ(t) = k= 1 x(k ) h (t k )
LTI System

Letting ! 0:
– (t) ! (t)
– x̂(t) ! x(t) (piece-wise approximation becomes exact)
R1
– ŷ(t) ! y(t) = 1 x(⌧ )h(t ⌧ )d⌧ (by the Riemann approximation of
the integral),
where h(t) is the response of the system to (t).

16
 Continuous-time LTI Systems

Define the impulse response of a CT system to be the response of the system

to x(t) = (t) and denote it by h(t).

Let h(t) be the impulse response of an LTI system. Then for any input
x(t), the output is
Z 1
y(t) = x(⌧ )h(t ⌧ )d⌧.
1

The above integral is called the convolution (integral) of x(t) and h(t).
def
Denote it by y(t) = x(t) ⇤ h(t).

17
 MAIN MESSAGE

For LTI systems, it is sufficient to know the impulse response h(t) (or h[n])
to know the response to any arbitrary input x(t) (x[n]).
In other words, we know everything about these systems if we know the
impulse response and hence, the notation:

x(t) y(t)
h(t)

or

x[n] y[n]
h[n]

Only catch: computing convolution is not easy.

18
 Example

Clicker Question: Which of the following systems is linear time-invariant

(LTI)?
a. y[n] = 3x[n] + 2
R1
b. y(t) = 1 2⇡x(⌧ ) (u(t ⌧) u(t ⌧ 1))d⌧
c. y(t) = x(2t)
d. y[n] = (x[2n])2
y(t) = x(t) ⇤ 2⇡ (u(t) u(t 1))
For the other systems:

L TI
A. No (incrementally linear ) Yes
B. Yes Yes
C. Yes No
D. No No

19
 How to compute convolution?

Fix time n
P
y[n] = 1 k= 1 x[k]h[n k]
As a function of k: h[n k] is h[k] flipped and shifted to the right by n:
3
h[k]
2
1
k
2 2 4
3 3 3
h[ k] h[1 k] h[2 k]
2 2 2
1 1 1
k k k
4 2 2 4 2 2 2 2 4

Output at time n: sample by sample multiply x[k] by h[n k] and then add
Example: What is the output of an LTI system with the impulse response
h[n] = n(u[n] u[n 3]) to the input x[n] = u[n] u[n 3]?

20
 Computing convolution: DT

2
x[k]
1.5
1
0.5
k
4 2 2 4

P1
We want to compute y[n] = k= 1 x[k]h[n k] for all n
Let n = 2:
a. imagine h[n k]
b. multiply h[n k] and x[k] point by point
c. add them up
d. the result is the output at time n (y[n])

1
y[n]
0.5
n
2 2
0.5

21
 Computing convolution: DT

2
x[k]
1.5
1
0.5
k
4 2 2 4

P1
We want to comput y[n] = k= 1 x[k]h[n k] for all n
For n < 1: no intersection of non-zero parts

1
y[n]
0.5
n
2 2
0.5

22
 Computing convolution: DT

2
x[k]
1.5
1
0.5
k
2 2 4

P1
We want to compute y[n] = k= 1 x[k]h[n k] for all n
Let n = 1:
a. imagine h[n k]
b. multiply h[n k] and x[k] point by point
c. add them up
d. the result is the output at time n (y[n])

3
y[n]
2
1
n
2 2

23
 Computing convolution: DT

2
x[k]
1.5
1
0.5
k
2 2 4

P1
We want to compute y[n] = k= 1 x[k]h[n k] for all n
Let n = 2:
a. imagine h[n k]
b. multiply h[n k] and x[k] point by point
c. add them up
d. the result is the output at time n (y[n])
3
y[n]
2
1
n
2 2

Continuing this, we get:

3
y[n]
2

1
n
2 2 4 6

24
 How to compute convolution in CT?

Given input x(t) and impulse response h(t)

The main idea: fix time t
R1
y(t) = x(t) ⇤ h(t) = 1 x(⌧ )h(t ⌧ )d⌧
As a function of ⌧ , x(⌧ ) remains the same
As a function of ⌧ , h(t ⌧ ): h(⌧ ) flipped and shifted to t
Output at time t: sample by sample multiply x(⌧ ) by h(t ⌧ ) and then add
integrate
Example: What is the output of an LTI system with the impulse response
h(t) = t(u(t) u(t 3)) to the input x(t) = u(t) u(t 3)?
To obtain the complete signal y, we need to repeat the above for every t!

25
 Properties of Convolution

Commutative: For any signals x(t) (or x[n]) and h(t) (or h[n]), x(t)⇤h(t) =
h(t) ⇤ x(t) (or x[n] ⇤ h[n] = h[n] ⇤ x[n]).
Proof DT:
1
X 1
X
n k!`
x[n] ⇤ h[n] = x[k]h[n k] = x[n `]h[`] = h[n] ⇤ x[n]
k= 1 `= 1

Proof of CT is similar.

26
 Properties of Convolution

Associative: For any three signals x[n], h1 [n], and h2 [n]: (x[n] ⇤ h1 [n]) ⇤
h2 [n] = x[n] ⇤ (h1 [n] ⇤ h2 [n]) and similarly: (x(t) ⇤ h1 (t)) ⇤ h2 (t) = x(t) ⇤
(h1 (t) ⇤ h2 (t)).
Distributive: For any three signals x[n], h1 [n], and h2 [n]: x[n] ⇤ (h1 [n] +
h2 [n]) = x[n] ⇤ h1 [n] + x[n] ⇤ h2 [n] and similarly: (x(t) ⇤ (h1 (t) + h2 (t)) =
x(t) ⇤ h1 (t) + x(t) ⇤ h2 (t)).
What is the system theoretic meaning? (serial and parallel interconnection)

27
 Impulse Response

x(t) y(t)
h(t)

We know that impulse response tells us everything about a system.

Can we determine the properties (memorlyess, causality, invertibility, BIBO
stability) of an LTI system by investigating h?
Answer: Yes! For LTI systems we can do it through investigating the impulse
response.

28
 Memoryless

x(t) y(t)
h(t)

A discrete-time LTI system is memoryless if and only if

h[n] = a [n], for some a 2 C.

A continuous-time LTI system is memoryless if and only if

h(t) = a (t), for some a 2 C.

29
 Causal

x(t) y(t)
h(t)

A discrete-time LTI system is causal if and only if

h[n] = 0, for n < 0. (1)

A continuous-time LTI system is causal if and only if

h(t) = 0, for t < 0.

30
 Invertibility

x(t) y(t)
h(t)

A discrete-time LTI system is invertible if and only if there exists a g[n] such that:

g[n] ⇤ h[n] = [n].

A continuous-time LTI system is invertible if and only if there exists a g(t) such
that:
g(t) ⇤ h(t) = (t).

31
 BIBO Stability

x(t) y(t)
h(t)

A discrete-time LTI system is BIBO stable if and only if:

1
X
|h[n]| < 1. (2)
n= 1

A continuous-time LTI system is BIBO stable if and only if:

Z 1
|h(t)|dt < 1.
1

32
 Practice Problem

Consider the DT - LTI system S with impulse response

✓ ◆n
1
h[n] = n u[n 1].
3
Which one of the following statements is true:
a. S is causal and stable.
b. S is causal but not stable.
c. S is not causal but is stable.
d. S is neither causal nor stable.

33
 Road map

NO CONVOLUTION!!!

34
 Key Idea

Suppose that we can find a rich family of signals vk (t) such that:

vk (t) k vk (t)
h(t)

for a scalar k 2C
And suppose that you can write a signal x(t) as:
1
X
x(t) = ak vk (t).
k= 1

Then what would be the response to x(t)?

P1 P1
x(t) = k= 1 ak vk (t) y(t) = k= 1 ak k vk (t)
h(t)

35
 Eigenfunctions

A signal (function) v(t) is an eigenfunction of a system if:

x(t) = v(t) y(t) = v(t)

System

for some 2 C.
Can we find an eigenfunction for an LTI system?
Let’s try v(t) = 1:
R1
x(t) = 1 y(t) = 1 h(⌧ )1d⌧
h(t)

R1
So, it is an eigenfunction with = 1 h(⌧ )d⌧

36
 Eigenfunctions

Is there any other eigenfunctions?

Conjecture: more generally x(t) = est where s 2 C
Let’s investigate:
R1 s(t ⌧ )
x(t) = est y(t) = 1 h(⌧ )e d⌧
h(t)

We have
Z 1 Z 1
s(t ⌧ )
y(t) = h(⌧ )e d⌧ = h(⌧ )est e s⌧
d⌧
1 1
Z 1
st s⌧
=e h(⌧ )e d⌧
1
R1 s⌧
H(s) = 1 h(⌧ )e d⌧ is called the transfer function of the system

Exponentials est are eigenfunctions of any CT LTI system for all s 2 C

such that H(s) exists. In particular, for x(t) = est the output will be
y(t) = H(s)x(t) = H(s)est .

x(t) = est y(t) = H(s)est

h(t)

37
 Example

Compute the transfer function of the CT integrator system. Using that,

compute the output of the system for the input x(t) = et + e4t .

38
 Road map: Discrete Time

Suppose that we can find a rich family of signals vk [n] such that:

vk [n] k vk [n]
h[n]

for a scalar k 2C
And suppose that you can write a signal x[n] as:
1
X
x[n] = ak vk [n].
k= 1

Then what would be the response to x[n]?

P1 P1
x(t) = k= 1 ak vk [n] y(t) = k= 1 ak k vk [n]
h[n]

A signal (function) v[n] is an eigenfunction of a system if:

x[n] = v[n] y[n] = v[n]

System

for some 2 C.
is often called an eigenvalue

39
 Eigenfunctions

Conjecture: x[n] = z n where z 2 C is fixed, is an eigenfunction

Let’s investigate:
P1 n k
x[n] = z n y[n] = k= 1 h[k]z
h[n]

We have
1
X 1
X
n k
y[n] = h[k]z = h[k]z n z k

k= 1 k= 1
X1
n k
=z h[k]z
k= 1

P1 k
H(z) = k= 1 h[k]z is called the transfer function of the system.

Exponentials z n are eigenfunctions of any DT LTI system for all z 2 C

such that H(z) exists. In particular, for x[n] = z n the output will be
y[n] = H(z)x[n] = H(z)z n .

x[n] = z n y[n] = H(z)z n

h[n]

Example: Compute the transfer function of the DT derivative (di↵erence)

system. Using that, compute the output of the system for the input x[n] =
cos(!n).

40