Digital Signal Processing

COMP 407

Madhav P Pandey*
* Compiled by

1
Signals and LTI systems in
frequency domain

2
Frequency Analysis of Signals
In addition to time domain representation, signal and
systems can also be represented and analyzed in frequency
domain.

In many situations and applications, such representations

are very useful and vital.

Several transformation tools are used to obtain such

frequency domain representations.

Fourier analysis tools including Fourier Series and

Transforms are one of the most important such tool.
3
 Fourier Analysis
Representation of signals as linear combination of
complex exponential (or alternatively sinusoidal)
signals.
The system output also then can be expressed as sum of
outputs to the individual component in the combination.
The resulting representation of systems is called the
frequency response.

Fourier Series are used to represent periodic signals

while Fourier Transforms are used to represent
aperiodic finite duration signals.
Fourier analysis tools are available for both CT and DT
signals/systems. 4
 Fourier Series

Successive approximation of common signals using Fourier Series.5

Fourier Series
Expressing any periodic signal as sum
of harmonically related sinusoidal
signals.

6
Fourier Analysis of DT Signals
Discrete time signals may be represented in terms of
sinusoids or complex exponentials.

Such representation is called Fourier representations.

Periodic signals are represented by Fourier Series while

aperiodic signals are represented by Fourier Transforms.

Importance of Fourier representations lie in the fact that

a wide range of practical signals have Fourier
representations. Also, complex exponentials are the eigen
functions for LTI systems.
7
Complex Exponentials & LTI Systems
Let us consider a DT LTI system with impulse response h[n].
If the input to the system is an complex exponential signal
x[n]=esn (where s is a general complex variable), the output y[n]
is given by the convolution as:
 
y[n]   h[k ]x[n  k ] 
k   
 h[
k  
k ]e s[ n  k ]

y[n]  e sn
 h[ k
k  
]e  sk


y[n]  x[n] H ( s ) where, H ( s )   h[ k
k  
]e  sk

Thus, the output of the LTI system for the complex

exponential signal as input is the same signal with only a
change in amplitude. H(s) is a complex constant whose value
is fixed for a known value of s.
8
Complex Exponentials & LTI Systems
In other words, all complex exponential signals esn (ie.,
for any value of s) are eigen function of the LTI system.
The corresponding eigen value associated with the
function being H(s).

Thus, the output of the LTI system for the complex

exponential signal as input is the same signal with only a
change in amplitude. H(s) is a complex constant whose
value is fixed for a known value of s.
In mathematics, an eigenfunction of a linear operator D defined on
some function space is any non-zero function f in that space that, when
acted upon by D, is only multiplied by some scaling factor called
an eigenvalue….. (wikipedia.com)
9
 DT Fourier Series
Consider a periodic complex exponential signal [n]  e j n
0

with fundamental frequency Ω0 and fundamental period

N=2π/Ω0. Also consider the set of complex exponential signals
2
jk n jk n
that have k [n]  e 0
e N
k  0,  1,  2,  3.........
Each of these signals have fundamental frequency that is
multiple of Ω0 and are called harmonically related complex
exponentials.
Since, the DT signals with frequencies spaced by 2π are
same, the set of harmonically related signals stated above has
only N distinct signals. That is, when k is changed by integer
multiples of N, same signals are generated
For a particular value of k,
k [n]  k  rN r  1, 2,3.........
10
 Fourier Series
The Fourier Series is the weighed linear combination of
harmonically related complex exponentials in the form:
2
jk n
x[n]   k
a
k N
e jk 0 n
 a e
k N
k
N

If any periodic signal x[n] can be written in the form as

above, it is called the Discrete Time Fourier Series
representation of the signal. The signal given as sum above is
also periodic with period N.
The term for k=0 is a constant. The terms for k=1 and k = -1
both have fundamental frequency Ω0 and are referred to as
fundamental components or first harmonics. The terms for k=2
and k=-2 both have fundamental frequency 2Ω0 and are
referred to as second harmonics. In general, k=P and k=-P
represent Pth harmonics. 11
 Fourier Series
The set of complex coefficients ak  is called Fourier Series
Coefficients or Spectral Coefficients of the signal x[n].
They provide a measure of portion of the signal x[n] that is at
each harmonic component.
For example, a0 gives the dc component (average value) of
x[n].
Similarly, a2 gives the component with frequency 2Ω0 and so
on..
Plot of ak  is known as spectrum of x(t).
In general FS coefficients may be complex. So, magnitude
spectrum and phase spectrum are required.
The range of values that k can take may be k=0,1,2,3…N-1 or
k=1,2,3…….N or k=-1,0,2,3….N-2 and so on….
12
 DT Fourier Series
A periodic DT signal can be expressed as weighed linear
combination of a complex exponential and its harmonics.
The resulting series is called Fourier series representation

For a periodic signal x[n] with fundamental period N and

frequency 0,
2
jk n
x[n]   k
a
k N
e jk0 n
 a e
k N
k
N

Where Fourier Coefficients ak are given as:

2
1  jk n
ak  
N n N
x[n]e N
k  0,1, 2....N  1
13
 F S Coefficients - An Example
# x[ n]  sin  0 n

As, already known, a DT sinusoidal signal sinΩ0n is periodic

only if 2 0 is an integer or ratio of integers. When it is integer,
it takes the form of 0  2 N. When it is ratio of integers, it takes
the form 0  2M N .
For 0  2 N 2 2
1 j N n 1 j N n
x[n]  e  e
2j 2j
Comparing it with the synthesis equation of FS,
1 1
a0  0, a1  , a1   , ak  0 for other k
2j 2j
Note that there are N distinct coefficients and FS coefficients
are periodic with period N. i.e., ak  ak  N 14
 F S Coefficients - An Example
2
For example if N=5, i.e., x[n]  sin n
The spectrum will look like: 5

For 0  2M N
If there are no common factors in M and N, the fundamental
period of x[n] will be N.
2 2
1 jM n 1  jM n
x[n]  e N
 e N
2j 2j
15
 F S Coefficients - An Example
Comparing it with the synthesis equation of FS
1 1
a0  0, aM  , a M   , ak  0 for other k
2j 2j
Note that there are N distinct coefficients and FS coefficients
are periodic with period N. 2
For example if N=5 and M=3, i.e., x[n]  sin 3 n
5
The FS spectrum looks like:

16
 DT Fourier Series -Properties
DTFS is a finite series. DTFS coefficients are periodic.

Let us consider x[n] and y[n] be periodic signals with period

N such that,
x[n]  ak
DTFS

y[n]  bk
DTFS

Linearity:
For constants A and B,
A x[n]  B y[n]  A ak  B bk
DTFS

17
 DT Fourier Series -Properties
Time shifting:
Shifting the signal in time results in change of phase
in spectrum.
 jk0 n0
x[n  n0 ]  ak e
DTFS

Frequency shifting:
e jM0 n x[n]  ak  M
DTFS

Time Reversal:
x[n]  a k
DTFS

Conjugation:
x*[n]  a * k
DTFS
18
DT Fourier Series -Properties
Periodic Convolution:
 x[
l N
l ] y[ n  l ]   N ak bk
DTFS

Multiplication:
x[n] y[n] 
DTFS
 ab
l N
l k l

First Difference:
x[n]  x[n  1]  
 1  e  jk0 ak
DTFS

Running Sum:

1

k  
x[k ] 
DTFS


1 e  jk0
ak
 19
 DT Fourier Transform
Discrete Time Fourier Series provides the representation of
DT periodic signals as linear combination of complex
exponentials.
This concept can be extended to apply for signals that are not
periodic. Resulting representation is called Discrete Time
Fourier Transform.
However, in contrast to FS, the complex exponentials
involved in the combination are infinitesimally close in
frequency and the sum converts to an integral.
The basic principle behind this development is same as that
we considered in development of CT Fourier Transform. i.e., an
aperiodic signal can be thought of as a periodic signal with an
infinite period.
20
 DT Fourier Transform
DTFT is used to represent aperiodic signals in terms of
sinusoids or complex exponentials.
For a finite duration signal x[n],

x[n] 
1
2  X e j
e jn
d
2

 
Where, X e j is called the Fourier Transform of x[n] and is
given as:
X e    x[n]e

j  j n

n  

DTFT is a continuous function of  and periodic in  with

period 2. 21
 Examples
Find DTFT of : x[n]  a n , a 1 For a>0

By definition,
 1 
j
X (e )   x[n]e
n  
 jn
  a  n  jn
e   e
a n  jn

n   n0

  ae    ae   1   ae  j 

   
 a e  j n j n
n jn n

n 1 n0 n 0 n 0

1 1 For a>0
  1 
1  ae j 1  ae  j
ae j 1
 j

1  ae 1  ae  j
1  a2

1  2a cos   a 2
22
Fourier Transform-Properties
Let us consider x[n] and y[n] be the finite duration signals
such that,
j
x[n]  X (e )
DTFT

y[n]   Y (e j )
DTFT

Linearity:
For constants A and B,
A x[n]  B y[n]   A X (e j )  B Y (e j )
DTFT

Time shifting:
Shifting the signal in time results in change of phase
in spectrum  jn j
x[n  n0 ]  e
DTFT 0
X (e )
23
Fourier Transform-Properties
Frequency Shifting:
e j0 n x[n]   X (e j ( 0 ) )
DTFT

Conjugation:
 j
x [n]  X (e
* DTFT *
)

Differencing and Accumulation:

x[n]  x[n  1]   (1  e  j ) X (e j )
DTFT

n 
1

k  
x[k ] 
DTFT
 j
(1  e )
X (e )   X (e )   (  2k )
j j0

k  

24
Frequency Response of systems
As already known, complex exponentials are eigen
functions of LTI systems. For LTI systems, output is given
by convolution of input and impulse response.
y[n]  x[n] * h[n]
Using Fourier Transform in both sides,
j j j
Y (e )  X (e ) H (e )
j
j Y ( e )
H (e ) 
X ( e j )
H (e j ) The Fourier Transform of h[n] which is also equal
to the ratio of Y (e j ) to X (e j ) is called frequency response.
25
Frequency Response of systems
Frequency response is used to represent LTI systems in
frequency domain.

H (e j ) is called the magnitude response while H (e j )

is called the phase response.
Y (e j )
H ( e j )  and H (e j )  Y (e j )  X (e j )
X ( e j )
Using Z-transform,
Y ( z)
Y ( z)  X ( z) H ( z) H ( z) 
X ( z)
H (z ) , The Z-Transform of h[n] which is also equal to the
ratio of Y (z ) to X (z ) is called the system function. 26
Complex Exponentials and LTI systems

h[n]
x[n]  e j0 n y[n]

The output is given by convolution:

  
y[n]   h[k ]x[n  k ] 
k  
 h[
k  
k ]e j 0  n  k 
  h[ k
k  
]e j 0 n  j 0 k
e

 e j0 n  h[ k
k  
]e  j 0 k
 e j0 n
H ( e  j 0
)

Thus, response to a complex exponential input is also a

complex exponential with same frequency but multiplied by
a factor H (e  j0 ) .
Complex Exponentials and LTI systems

h[n]
x[n]  e jn
y[n]  x[n]H e jn 

The complex exponential thus is the Eigen function of

the LTI systems while the Eigen value associated with it is :

H ( e  j )   h[ k
k  
]e  jk

Thus, if a input signal can be expressed as sum of

complex exponentials, output can then be computed as
same sum scaled by the Eigen values.
Complex Exponentials and LTI systems
Determine the output of the LTI system if
h[n]   1 n
2  u[n] x[n]  Ae
j 2 n

Here,
  e 
 
1
H (e  j
)   e
1 k  jk 1  j k

k 0
2
k 0
2
1  12 e  j
At   2 ,
 j 2 1 1 2  j 26.6
H (e )  j 2
  e
1 e
1
2
1 2 j
1
5
So,
 2  j 26.6  j 2 n 2 j  2 n  26.6 
y[n]  A e  e  Ae
 5  5
Complex Exponentials and LTI systems
What will be the output of the same system if
x[n]  10  5 sin 2 n  20 cos n
Here,
x[n]  10  5 
j n1
2j e
 j n 2jn
 21j e 2
 20  e
1
2  12 e  jn 
Now, 2 j 26.6 H (e j 2 )  2 e  j 26.6
 j 2
H (e  j 0 )  2 H (e )  e
5 5
j
H ( e  j )  2
3
H (e )  23
So,
 2  j 26.6 j 2 n 2 j 26.6  j 2 n   1 2 jn 1 2  jn 
y[n]  10  2  5 e e  e e   20   e   e 
2 j 5 2j 5  2 3 2 3 

40
y[n]  20  2 5 sin  2 n  26.6  cosn 

3
 LCC Difference Equations
An important class of LTI systems are those for which
the input and output satisfy an Nth order linear constant
coefficient difference equations of the form:
N M

a
k 0
k y[n  k ]   bk x[n  k ]
k 0

Taking z-transform on both sides,

N M

 k
a
k 0
z k
Y ( z )   k X ( z)
b z k

k 0 M

Y ( z) k
b z k

H ( z)   k 0
N
X ( z)
 k
a
k 0
z k
Frequency Response of systems
For a system described by the difference equation,
N M

a
k 0
k y[n  k ]   bk x[n  k ]
k 0
The frequency response is given as:
M

k
b e  jk

H ( e j )  k 0
N

 k
a e
k 0
 j k

In terms of poles and zeros,

 1  c e 
M
 j
k
j b0
H (e )  k 1

 1  d e 
N
a0  j
k
k 1 32
Frequency Response of systems
The magnitude response is then,

 1  c e 
M
 j
k
j b0
H (e )  k 1

 1  d e 
N
a0  j
In dB, k 1
k

  20 log10 1  ck e    20 log10 1  d k e  j 
M N
j b0  j
20 log10 H (e )  20 log10
a0 k 1 k 1

Attenuation in dB,
A  20 log10 H (e j )
Phase Response:
H (e )     1  ck e    1  d k e  j 
j b0 M  j
N

a0 k 1 k 1

Group Delay: gradient of H (e j )

33