EEEN 321

Signals and Systems

Lecture Notes 03

1
EEEN 321
 System response & convolution
• Let h[n] denote the response of a system to the unit impulse δ[n]:

h[n]
δ[n]
System
n n
0 0

• The linearity and time-invariance properties imply that:

x[k] h[n-k]
x[k] δ[n-k]
LTI System
n n
0 k 0 k

• In general:
 
 x[k ] [n  k ] LTI system  x[k ]h[n  k ]
k  k  

 x[n]  y[n]
2
EEEN 321
 System response & convolution

• In general, the output of a linear time-invariant (LTI) system is obtained as

the convolution of the input signal with the system impulse response h[n]:

LTI system
x[n] y[n]
h[n]


y[n]   x[k ]h[n  k ]
k 

• Alternative notation for convolution

y[n]  x[n]  h[n]

• This expression specifies the output y[n] of an LTI system for any input x[n]
in terms of the system impulse response h[n]
• Hence the impulse response h[n] fully characterizes the system
3
EEEN 321
 Visualization of convolution
LTI system
x[n] h[n] y[n]
1
0.5 2
0.25 n 1 1
n
? ? n
0.5


y[n]   x[k ]h[n  k ]
k 
   
 y[2]   x[k ]h[2  k ] y[1]   x[k ]h[1  k ] y[0]   x[k ]h[0  k ] y[1]   x[k ]h[1  k ] 
k  k  k   k 

Step-by-step procedure:
1. Select n (which is now fixed), multiply x[k] and h[n-k] for all k and add to
obtain y[n].
2. For n+1, shift h[n-k] by one to the right and multiply with x[k] and add.
3. Etc. For n+2, n+3,...
4
EEEN 321
 Visualization of convolution (2)
LTI system x[k]
y[n] 1
x[n] h[n] 0.5
1 k
0.5
1
n
? ? n
n
h[-1-k]
 1
y[n]   x[k ]h[n  k ] y[1]  0
1
k
k  

h[-k]
2
1
y[n] k y[0]  11  1

1.5
h[1-k]
1 3 1
0.5
k y[1]  11  1 0.5  1.5
n
-2 -1 0 1 2 3 4

4 h[2-k]
1 y[2]  1 0.5  0.5
k

5 h[3-k]
1 y[3]  0
k
5
EEEN 321
 Visualization of convolution (3)
x[k] x[k]
1
0.5
5 1
0.5
0.25 k 0.25 k

1 h[-4-k] 6 h[-k]
y[0]  0.25 1  1 2  0.5 1  2.75
2 2
1 1
1 y[4]  0.25  0  1 0  0.5  0  0 1
0.5 k 0.5 k

2 h[-3-k] 7 h[1-k]
2
1
2
1 y[1]  0.25  0  11  0.5  2  2
1
0.5 y[3]  0.25  0.5  1 0  0.5  0  0.125 1
k
k 0.5

3 2
h[-2-k] 8 h[2-k] 2 y[2]  0.25  0  1 0  0.5 1  0.5
1 1
1
0.5
y[2]  0.25 1  1 0.5  0.5  0  0.75 1
k 0.5 k

4 h[-1-k] h[3-k]
2 9 2 y[3]  0.25  0  1 0  0.5  0  0
1
1 y[1]  0.25  2  11  0.5  0.5  1.75 1
1
0.5 k 0.5 k

6
EEEN 321
 Visualization of convolution (3)

y[n]
2.75

LTI system 1.75

2

x[n] h[n]
1
0.5 2 0.75
0.5
0.25 n 1 1
0.5 n 0.125 n

7
EEEN 321
 Sifting property of the cont.-time unit impulse
• A sifting property also holds for the continuous-time impulse – similar to that
established for the discrete-time impulse
 
x[n]   x[k ] [n  k ] x(t )   x( ) (t   )d
k  
Sifting property of δ[n] Sifting property of δ(t)

• Idea of the proof of the sifting property

Staircase approximation of x(t):
x(t )

  (t ) x(k)      (t  k)
x (t )   x(k)      (t  k)
k 
1/Δ x (t )
x(k)  x( )
  (t  k)   (t   )
As Δ → 0:
t t   d

0
Δ
kΔ   
 
Therefore: x(t )  lim x (t )  lim
 0

  0 k 
x(k)      (t  k)  x(t )   x( ) (t   )d

8
EEEN 321
 System response & convolution
• Let hΔ(t) denote the response of a system to a narrow pulse δΔ(t):

ΔhΔ(t)
Δδ(t)
1 System
t t
0 Δ 0

x(0)ΔhΔ(t)
x(0)Δδ(t)
x(0)
Linear
t system t

0 Δ 0

x(kΔ)ΔhΔ(t-kΔ)
x(kΔ)Δδ(t-kΔ) Linear
time-invariant
x(kΔ) t system t

0 kΔ 0 kΔ

9
EEEN 321
 System response & convolution (2)
• Consider the addition of all input pulses (staircase approximation of x(t):

 x(k)      (t  k) 
x (t ) k   y (t )   x(k)    h (t  k)
k 
Linear
time-invariant
system
t t

0
 
lim x (t )  lim
 0   0 k 

x(k)      (t  k)  lim y (t )  lim
 0   0 k 

x(k)    h (t  k)


Therefore: x(t )  y (t )   x( )h(t   )d


x(t )
Linear y(t )
time-invariant
system
t t

0
10
EEEN 321
 System response & convolution (3)

• In general, the output of a linear time-invariant (LTI) system is obtained as

the convolution of the input signal with the system impulse response h(t):

LTI system
x(t ) y(t )
h(t)


y (t )   x( )h(t   )d

• Alternative notation for convolution

y(t )  x(t )  h(t )

• This expression specifies the output y(t) of an LTI system for any input x(t) in
terms of the system impulse response h(t)
• Hence the impulse response h(t) fully characterizes the system
11
EEEN 321
 Visualization of convolution (cont. time)
LTI system
x(t) y(t) 
h(t)
2
t
1
t ? ? t
y (t )   x( )h(t   )d
0 1
2

0 1
-1

• To compute y(t), view x and h as two functions of τ : x(τ) and h(τ)

• Flip (time reverse) the function h: h(-τ)
h(-τ)
1
τ
-1 0

• Increasing t amounts to shifting h from left to right (x remains fixed, not shifted)
h(t-τ)
1
t>0
τ
-1 0 t

• For each t, multiply x by the shifted h and compute the integral

12
EEEN 321
 Visualization of convolution (2)
x( )
2
2 3

0 1
-1

h(t   )
1 t  0  x( )h(t  )  0  y(t )  0

t

h(t   ) t
1
 
0  t  1  y (t )  2 (  1  t )d  t (2  t )
0
t

h(t   ) 1 t
3 2 11
1
 1  t  2  y (t )   2 (  1  t )d  1(  1  t )d  2 t  6t 
2
t 1 1
t-1 t

h(t   ) 2
1 9

2  t  3  y (t )   1 (  1  t )d   t 2  3t 
1
 2 2
t 1
t-1 t
h(t   ) y(t )
1 1
3  t  y(t )  0
 2 3 t
t-1 t 1
- 0.5 13
EEEN 321
 Visualization of convolution (3)

LTI system
x(t ) h(t) y(t )
2 1 1
2 3 t t 2 3 t
0 1 0 1 1
-1 - 0.5

14
EEEN 321
 Summary
• An LTI system is completely characterized in terms of its response to the
unit impulse (discrete-time or continuous-time unit impulse)
• The response of an LTI system to any input can thus be computed from the
impulse-response h[n] or h(t) of the system:


y[n]   x[k ]h[n  k ] y[n]  x[n]  h[n]
k 


y (t )   x( )h(t   )d y(t )  x(t )  h(t )


• Since h[n] or h(t) completely determines the behavior of an LTI system, the
properties of LTI systems can be expressed in terms of the impulse
response

15
EEEN 321
 Properties of LTI sytems
• Commutativity: xh  h x
+
h1
• Distributivity: x  (h1  h2 )  x  h1  x  h2 = h1+h2
+
h2

• Associativity: x  (h1  h2 )  ( x  h1)  h2  x  h1  h2

h1*h2 = h1 h2 = h2 h1

• Invertibility: given h, g such that h  g   x h g yx

 
• Causality: y[n]   x[k ]h[n  k ]
k 
y[n]   x[n  2]h[2]  x[n 1]h[1]  x[n]h[0]  x[n  1]h[1]  x[n  2]h[2]  

= 0 for causality
Causal system: h[n]  0, n  0 (discrete time)

h(t )  0, t  0 (continuous time)

16
EEEN 321
 Properties of LTI systems (2)
• Stability

A system is bounded-input bounded-output (BIBO) stable if and

only if its impulse response is absolutely summable (integrable)
 
 h[n]    h(t ) dt  
n   
(discrete time) (continuous time)

• Memorylessness

h[n]  A [n] h(t )  A (t )

(discrete time) (continuous time)

17
EEEN 321
 The unit-step response
Discrete time Continuous time

impulse impulse impulse impulse

 [n] h[n]
response
 (t ) h(t )
response
function LTI function LTI
unit step system unit step unit step system unit step
u[n] s[n] u(t ) s(t )
function response function response

s[n]  u[n]  h[n] s(t )  u(t )  h(t )

n t
  [k ]   ( ) d
d
u[n]   [n]  u[n]  u[n 1] u (t )   (t )  u (t )
k  
dt


n t
 
d
s[n]  h[k ] h[n]  s[n]  s[n  1] s (t )  h( ) d h(t )  s(t )
k 
dt


The unit-step response can also be used to characterize an LTI system

18
EEEN 321
 The unit-step response
Consider the following system
x(t) y(t )
2 LTI 3
system
0 t 0 1 2 t

What is the impulse response of the system?

Since 0.5y(t) is the system unit-step response, its impulse response is given by
d
h (t )  s (t ) with s(t )  0.5 y(t )
dt
h(t)
Hence
1.5
1 2
0 t
-1.5

19
EEEN 321
 Differential & Difference Equations
• An alternative approach is to relate the input and output signals through an
equation or set of equations (this is closely related to our systems discussions
so far)
• For continuous-time systems: linear constant-coefficient differential equations
• For discrete-time systems: linear constant-coefficient difference equations
Car motion: Bank account:

applied force
c[n] : capital at time n
d[n] : deposit at time n
frictional force r: interest rate

dv(t )  1
 v(t )  f (t ) c[n]  (1  r)c[n 1]  d[n]
dt m m

• These equations provide an implicit specification of the system  the

equations need to be solved
• In general, the solution requires that auxiliary conditions be satisfied
20
EEEN 321
 Solving linear const.-coeff. differential equations
dy(t )
• First order differential equation: a1  a0 y(t )  b0 x(t )
dt
• The complete solution is the sum of a particular and a homogeneous solution:
y (t )  y p (t )  yh (t )

y p (t ) : a particular solution for given input x(t)

yh (t ) : the solution for zero input x(t)

• Example: the input x(t )  3e2t for t ≥ 0 and x(t )  0 for t < 0 is given
• For a particular solution, assume that y(t) has a form similar (forced) to x(t)
y p (t )  Be2t , t  0
dy(t )
• The homogeneous solution (i.e., solution of a1  a0 y(t )  0 ) is of the type
dt
yh (t )  Aet
• Determine the unknown coefficients by using (given) auxiliary conditions

for example (initial rest condition): y(t  0)  0

21
EEEN 321
Solving linear const.-coeff. differential equations

Problem:

Find the solution of the differential equation

dy (t )
 3 y (t )  x(t )
dt
given that x(t )  3e 2t , t  0 and the system is at rest initially.

22
EEEN 321