Chapter 2 Systems and Signals

 Introduction
 Discrete-Time Signals: Sequences
 Discrete-Time Systems
 Properties of Linear Time-Invariant Systems
 Linear Constant-Coefficient Difference Equations
 Frequency-Domain Representation of Discrete-Time Signals
and Systems
 Representation of Sequence by Fourier Transforms
 Symmetry Properties of the Fourier Transform
 Fourier Transform Theorems
 1. Introduction-- Definition
2

 Signals
 Any physical quantity that varies with time, space or any other
independent variable.
 Communication beween humans and machines.
 Systems
 mathematically a transformation or an operator that maps an
input signal into an output signal. S
 can be either hardware or software.
 such operations are usually referred as signal processing.
 Digital Signal Processing
 The representation of signals by sequences of numbers or symbols
and the processing of these sequences.
 1. Introduction-- Definition
3
 1. Introduction-- Classification
4

 Continuous-Time versus Discrete-Time Signals

 Continuous-time signals are defined for every value of time.
 Discrete -time signals are defined at discrete values of time.
 Continuous-Valued versus Discrete-Valued Signals
 A signal which takes on all possible values on a finite range or infinite
range is said to be a continuous-valued signal.
 A signal takes on values from a finite set of possible values is said to be
a discrete-valued signal.
 Multichannel versus Multidimensional Signals
 Signals may be generated by multiple sources or multiple sensors. Such
signals are multi-channel signals.
 A signal which is a function of M independent variables is called
multi-dimensional signals.
 1. Introduction-- Basic Elements
5

 A/D Converter Analog Digital

Input Input
 Converts an analog Signal Signal
signal into a A/D
Converter
sequence of digits
Digital
 D/A Converter 0
t {3, 5, 4, 6 ...}
Signal
Processing
 Converts a sequence D/A
of digits into an Converter

analog signal Analog Digital

Output Output
Signal Signal
 1. Introduction-- Advantages of Digital over Analog
Processing
6

 Better control of accuracy

 Easily stored on magnetic media
 Allow for more sophisticated signal
processing
 Cheaper in some cases
 Examples
7

 A picture is a two-dimensional signal

 I(x,y) is a function of two variables.
 A black-and-white television picture is a three-
dimensional signal
 I(x,y,t) is a function of three variables.
 A color TV picture is a three-channel, three-dimensional
signals
 Ir(x,y,t), Ig(x,y,t), and Ib(x,y,t)
...............
 2. Discrete-Time Signals: Sequences
8

 Continuous signals  Linear combination form



 f(x), u(x), and s(t), and so on. x [ n ]   x [ k ] [ n  k ]


k

 Sampled signals  Unit step sequence

 s[n], f[n], u(n). 0 , n  0
u[ n ]  
 Shifting by k units of time 1 , n  0

y[n]  x[n  k ]  Sinusoidal Function

sin( n   )  sin( 2fn   )
 Unit sample sequence
xa(t)
(Kronecker Delta) t
0 , n  0 Time
 (n)  
1 , n  0 x [n] n
 2. Discrete-Time Signals: Sequences (c.1)
9

 Definition
Xa(t) = A cos( W t+ ), - ∞ 
<t<∞
 A is the amplitude of the sinusoid.
 W is the frequency in radians per second. Time
  is the phase in radians.
 F=W/2 is the frequency in cycles per second or hertz.
 Remarks
 The fundamental period is T=1/F.
 For every fixed value of F, f(t) is periodic
– f(t+T) = f(t), T=1/F
 Continuous-time sinusoidal signals with distinct frequencies are themselves
distinct.
 Increasing the frequency F results in an increase in the rate of oscillation.
 2. Discrete-Time Signals: Sequences (c.2)
10

 Discrete-Time Sinusoidal Signals

 X(n) = A cos( n+  ), n =1, 2, ...
 A is the amplitude of the sinusoid

  is the frequency in radians per sample

  is the phase in radians

 f=/2 is the frequency in cycles per sample.

X(n) = A cos(  n+ )
 2. Discrete-Time Signals: Sequences (c.3)
11

 Discrete-Time Sinusoidal Signals X(n) = A cos(  n+ )

 A discrete-time sinsoidal is periodic only if its
frequency f is a rational number
– X(n+N) = X(n), N=p/f, where p is an
integer
 Discrete-time sinusoidals where frequencies are
separated by an integer multiple of 2 are identical
– X1(n) = A cos( 0 n)
– X2(n) = A cos( (0 2) n)
 The highest rate of oscillation in a discrete-time
sinusoidal is attained when = or (=-), or
equivalently f=1/2.
– X(n) = A cos(( 0+)n) = -A cos((0+)n
 2. Discrete-Time Signals: Sequences (c.4)
12

 Exercise: Find the periods.

 sin0.1k, cos 10.1k, sin0.1k, cos3k/7
 An observation
 For the sampling process f[k]= f(kT), the mapping between discrete-frequency
and analog-frequency is one-to many.

 Example
– sin1.1t, sin3.1t, sin3.1t, sin(-0.9)t, sin5.1pt, sin (-2.9)t have the sample digital
envelop for sampling time T=1.
– Reason : 1T -2T = nx(2)
 2. Discrete-Time Signals: Sequences (c.5)
13

 The Digital Frequency

 If 1T and 2T differ by a multiple of 2, they are still
considered equal.
 The digital frequency can be restricted to lie in an interval of 2
to eliminate this nonuniqueness.
(-, ], [0, 2) or [, 3]
 If (-, ] is selected, then Digital
/T
3/Ts /Ts /Ts 3/Ts 5/Ts
Analog
0 
/T
 Example: Find the frequencies of the following sequences for T=1
– sin 4.2k, cos5.2k, sin 10k, sin(-2.1k), cos20k
 3. Discrete-Time Systems
14

 Systems
 Mathematically a transformation or an operator that maps
an input signal into an output signal
 Can be either hardware or software.

 Such operations are usually referred as signal processing.

 E.x. n n
y( n )   x(k )   x(k  1)  x(n)  y(n  1)  x(n)
k  k 

n n
Discrete-Time
System H
 3. Discrete-Time Systems-- Classification
15

 Time-Invariant versus Time-Variant Systems

 A system H is time-invariant or shift invariant if and only if
 x(n) ---> y(n)
implies that
 x(n-k) --> y(n-k)
 for every input signal x(n) x(n) and every time shift k.
 Causal versus Noncausal Systems
 The output of a causal system satisfies an equation
 y(n) = F[x(n), x(n-1), x(n-2), ...].
 where F[.] is some arbitrary function.
 Memory versus Memoryless Systems
 A sysetm is referred to as memoryless system if the output y(n) at every
value of n depends only on the input x(n) at the same value of n.
 3. Discrete-Time Systems-- Classification
16
(c.1)
u1(n)
 Linear versus Nonlinear Systems a
Linear y(n)
 A system H is linear if and only if + systems
 H[a1x1(n)+ a2 x2 (n)] = a1H[x1 (n)] + a2H[x2 (n)] b
u2(n)
 for any arbitrary input sequences x1(n) and x2(n),
and any arbitrary constants a1 and a2.
 Multiplicative or Scaling Property
 H[ax(n)] = a H[x(n)] u1(n) Linear
systems a
 Additivity Property y(n)
+
u2(n) Linear
H[x1(n) + x2 (n)] = H[x1 (n)] + H[x2 (n)] systems b
 Stable versus Unstable Systems
 An arbitrary relaxed system is said to be bounded-
input-bounded-output (BIBO) stable if and only if
every bounded input produces a bounded output.
 4. Linear Time-Invariant Systems
17

 The Importance of LTI Systems

 Powerful analysis techniques exist for such systems.
 Many real-world systems can be closely approximated for linear,
time-invariant systems.
 Analysis techniques for LTI systems suggest approaches for that of
nonlinear systems.
 Linear Systems
 Time-Invariance
 y(n)=H[x(n)] ==> H[x(n-k)] = y(n-k)
 Linear superposition
Impulse Response
 H[a1x1(n)+a2x2(n)] = a1H[x1(n)]+a2H[x2(n)]
of the System
= a1y1(n) + a2y2(n)
 Specification for LSI Systems 
H [ x ( n )]  H [
k
 x ( k )  ( n  k )]   x ( k ) H [  ( n  k )]   x ( k ) h ( n  k )
 
k 
k
 4. Linear Time-Invariant Systems
18
 4. Linear Time-Invariant Systems (c.1)
19

 Linear Systems (c.1)

 Finite Impulse Response (FIR)or Infinite Impulse Response (IIR)
 Depends on the the finite and infinite number of terms of h(n)
 Convolution Formula

y ( n )  x ( n ) h ( n ) 
k
 x (k )h(n  k )


 Ex.
 A saving account with monthly interest rate 0.5%. The interest is added to the principal at the
first day of each month. If we deposit u[0] = $100.00, u[1] = -$50.00, u[2] = 200.00 what is
the total amount of money. What is the total amount of money on the first day of the fifth
month ?
 y[0] y[1] y[2] y[3] ...

4. Linear Time-Invariant 1 2 3 4 -1

Systems (c.2)
20
-1
1
-1
1
-2 -3
2 3
-4
4
1
-1

 Discrete Convolution-- Table Lists -3 -3 -6 -9 -12 3

 Consider the convolution of 2 2 4 6 8 -2

{1, 2, 3, 4, -1, ...} and {-1, 1, -3, 2, ...}

 Discrete Convolution--
Graphical Computation
 Flipping h [i] to yield h [-i]
 Shifting h [-i] to yield h [k-i]

 Multiplication of h [k-i] and u [i]

 Summation of h [k-i]u [i] from i =0, 1, 2, ...

k k
y[ k ]   h[k  i]u[i]   h[i]u[k  i]
i0 i0
 5. Properties of the LTI Systems
21

x[n] y[n]
 Communicative 
h[n]
y [k ]   h[k  i]u[i]  h[k ] * u[k ]
i  
h[n]
x[n]
y[n]

  h[i]u[k  i]  u[k ] * h[k ]
i  
x[n] y[n]
h1[n]+h2[n]
 Parallel Sum
x[n] y[n]
x[n]*{h1[n]  h2 [n]} h1[n] +
 x[n]* h1[n]  x[n]* h2 [n]
h2[n]
y[n]
 Cascade Form x[n]
h1[n] h2[n]

x[n] *{h1[n] * h2 [n]} x[n] y[n]

 {x[n] * h1[n]} * h2 [n] h1[n]*h2[n]
 5. Properties of the LTI Systems (c.1)
22

 Stability of LTI Systems (BIBO, Bounded-Input-Bounded Output System)

 Linear time-invariant systems are stable if and only if the impulse
response is absolutely summable, i.e., if

S 
k
 h[k ]  
 

<pf> q -> p
~q -> ~p
If S=  then the bounded input
Since that

y[n]   h[k ]x[n  k ]
k  
 h *[  n ]
 , h[n ]  0

x [ n ]   h [ n ]
 0,
  h[k ] x[n  k ]
k  
 h[n ]  0

 If x[n] is bounded so that x [ n ]  Bx will generate

 then   
2
h[ k ]
y [ n ]  Bx 
 
k
h[ k ] y [0] 
k
 x [ k ] h [ k ]  
 
k h[ k ]
S
 5. Properties of the LTI Systems (c.2)
23

 Ex. Find the impulse response of the following systems

y[ n]  x[ n  nd ]

M2
1
y[ n]   x[ n  k ]
M 1  M 2  1 k  M1

n
y[ n]   x[n]
k  

y[ n]  x[ n  1]  x[ n]
 6. Linear Constant-Coefficient Difference
24
Equations
 Input-Output Description
N M
y( n)    ak y( n  k )   bk x ( n  k )
k 1 k 0

 From Difference Equation to Impulse Response

 Not every convolution can be transformed into a simple
difference equation
 If the difference equation description of a system is known,
then the impulse response of the system can be readily
obtained.
Example y[k  1]  1005
. y[k ]  u[k  1]
or y[k ]  1005
. y[k  1]  u[k ]
 6. Linear Constant-Coefficient Difference
25
Equations
 Ex 2.16
 6. Linear Constant-Coefficient Difference
26
Equations (c.1)
 Solution Specification for the LCCDE
 The solution can be obtained recursively for either positive
time or negative time.
 Linear, time-invariant, and causal ==> the solution is unique.

 Initial Condition
 Initial-rest conditions if the initial condition is zero.
 Initial rest condition ==> LTI and causal

x(n) Linear Constant y(n)

Coefficient Equations
by {ak, bk, N, M}
 7. Frequency-Domain Representation of
27
Discrete-Time Signals and Systems
 Eigenfunctions of LTI Systems
 Response to Exponential Functions
jn  
 
y( n)  
k  
h( k )e j ( n  k )
e  
 k  
h( k )e jk   H ( e j )e jn

 Eigenvalues of the System Ex.
y [ n ]  x [ n  nd ]
 Frequency Response-- Complex M2
1
H ( e j )  H R ( e j ) jH I ( e j ) y[n]  x[n  k ]
M 1  M 2  1 k  M 1

 Magnitude and Phase X(n) = A cos(  n+  )

j
H ( e j )  H ( e j ) e jH ( e )
 7. Frequency-Domain Representation of
28
Discrete-Time Signals and Systems
 Ex. 2.17
 Method 1: Frequency Response

 Method 2: Derive from Impulse Response.

29
 8. Representation of Sequences by Fourier
30
Transforms
 Fourier Transform Pair
 Transform  Question:
X ( e j )   x ( n ) exp
 
n
 jn

1. Periodic Functions for

1 
x (n) 
2   X ( e j ) e jn d

the Transform.
2. Relationship with the
Frequency response.
 Inverse Transform (Synthesis Formula) 3. Inverse of each other ?
j 4. Existence of the
X ( e j )  X ( e j ) e jX ( e )
transform for functions.
 Magnitude & Phase Spectrum
 8. Representation of Sequences by Fourier
31
Transforms
 Proof of Fourier Pair
 Subsisting analysis equation into synthesis equation
 L'Hôpital's rule
32

 In its simplest form, l'Hôpital's rule states that for

functions ƒ and g:
 Existence of Fourier Transform
33

 Convergence of the infinite sum

The series can be shown to

converge uniformly to a
continuous function of 
 Existence of Fourier Transform
34

 Existence of Fourier transform of infinite sequence

 Existence of Fourier Transform
35

 Relax of the condition of uniform convergence of infinite

sum
 Existence of Fourier Transform
36

 Example
 8. Representation of Sequences by Fourier
37
Transforms (c.1)
 Example
 Lowpass Filters
j
 1,    c
H (e )
0 ,  c    

 Impulse Response
1  jn
2 
h[n ]  e d
sin  c n
 ,  n  
n

Causal ?
Decay Factor ?
Absolutely Summable ?
Gibbs phenomenon.
 8. Representation of Sequences by Fourier
38
Transforms
 8. Representation of Sequences by Fourier
39
Transforms (c.2)
 Fourier Transform of the periodic train 

x ( n )  1 for all n X (e j
)  2 (   2r )
r  


X (e j
) 

r
2 (    0  2r )
Absolutely Summable ?
Square Summable ?

1 
x[n] 
2   r 2 (   


0  2r )e jn d

1 
  2 (   
j 0 n
 ) e jn d  e for any n .
2  0

 



n
e j 0 n
e  jn  

r
2 (    0  2r )

Lighthill, 1958
 8. Representation of Sequences by Fourier
40
Transforms
 Fourier transform of sinusoidal signal
 



n
e j 0 n
e  jn
 

r
2 (    0  2r )

 Extend the theory

 9. Symmetry Properties of the Fourier
41
Transform
 Definition
 Conjugate-symmetric sequence
 xe[n]=x*e [-n]

 Conjugate-antisymmetric sequence
X ( e j )  X e ( e j ) X o ( e j )
 xo[n]=-x*o [-n]
1
 Properties X e ( e j ) 
2
[ X ( e j )  X *( e  j )]

1
A sequence can be represented as X o ( e j )  [ X ( e j )  X *( e  j )]
2
x[n] = xe[n] + xo[n]
X e ( e j )  X e* ( e  j )
1
xe [ n ]  ( x [ n ] x *[  n ]) X o ( e j )   X o* ( e  j )
2
1
x o [ n ]  ( x [ n ] x *[  n ])
2
 9. Symmetry Properties of the Fourier
42
Transform (c.1)
 Example
 Real sequence
 x[n]=anu[n]


X (e j
)  x ( n ) exp
 
n
 jn

1 
x (n) 
2   X ( e j ) e jn d

 9. Symmetry Properties of the Fourier
43
Transform (c.2)
 9. Symmetry Properties of the Fourier
44
Transform (c.3)
 10. Fourier Transform Theorems
45
 Theorem Examples
46
11. Discrete-Time Random Signals
 Precise description of signals are so complex
 Modeling the signals as a stochastic process.
 Let the means of the input and output processes are
mx [n]   {x[n]}, my [n]   { y[n]}

 If x[n] is stationary, then mx[n] is independent of n and

mx   {x[n]}, my   { y[n]}
 
my   { y[n]}   h[k ] x[n  k ]  m  h[k ]
k  
x
k  
11. Discrete-Time Random Signals
 Autocorrelation function of the output process
   
 yy [n, n  m]   { y[n] y[n  m]}      h[k ]h[r ]x[n  k ]x[n  m  r ]
k   r   
 
  h[k ]  h[r ] {x[n  k ]x[n  m  r ]}
k   r  

 If x[n] is stationary
 
 yy [n, n  m]   yy [m]   h[k ]  h[r ]
k   r  
xx [m  k  r ]
chh[l ] 


 h[k ]h[l  k ]
k  

 Let l=r-k
  
 yy [n, n  m]   yy [m]   h[k ]  h[l  k ]
k   l  
xx [m  l ]  
l  
xx [m  l ]chh[l ]
11. Discrete-Time Random Signals
 Fourier Transform
2
 yy (e j )  Chh (e j ) xx (e j ) Chh (e j )  H (e j ) H * (e j )  H (e j )

 Power density spectrum

 Cross-correlation
 
 
xy [m]   {x[n] y[n  m]}    x[n]  h[k ]x[n  m  k ]   h[k ]xx [m  k ]
 k    k 

 xy [e j ]  H (e j ) xx (e j )
11. Discrete-Time Random Signals
 Ex. White Noise
xx [m]   x2 [m]
 Power Spectrum

 The average power

 11. Concluding Remarks
51

 Introduction
 Discrete-Time Signals: Sequences
 Discrete-Time Systems
 Properties of Linear Time-Invariant Systems
 Linear Constant-Coefficient Difference Equations
 Frequency-Domain Representation of Discrete-Time Signals
and Systems
 Representation of Sequence by Fourier Transforms
 Symmetry Properties of the Fourier Transform
 Fourier Transform Theorems