UNIVERSITY TENAGA NASIONAL

DEPARTMENT OF ELECTRICAL AND ELECTRONICS ENGINEERING

Signals and Systems

Part IV

Continuous Complex Exponential Fourier Series

Discrete Time Complex Exponential Fourier Series

COMPILED BY WONG HUNG WAY

August 2010

1
Fourier Series

If given a continuous time signal x(t), and if x(t+T) = x(t), then the signal x(t) is said to be a
periodic signal, with period T.

2
The signal complex exponential x(t )  A exp j0t is periodic with period T  , because
0
x(t  T )  A exp j0 (t T )  A exp j0t exp j0T .

2
If T  then x(t  T )  A exp j0t exp j 2 and since exp j 2  cos 2  j sin 2  1
0

x(t  T )  A exp j0t  x(t )

Basic Idea of Fourier Series

2
xk (t )  exp jkot are harmonic signals with period Tk  , k0
k 0

Basic Fourier series idea : Represent all periodic signals x(t) as a series of the form

x(t )  ae
k 
k
jk0t

where ak are the Fourier Series Coefficients (Spectral coefficients) of x(t) and k are positive
integers.

k= 0 gives the DC signal component of x(t) i.e a0 provides information on the DC value of x(t).

k   1 yields the fundamental frequency or the first harmonic 0

k  2 gives the harmonics

2
 
Thus are two ways to look at a signal x(t) i.e either in the time domain x(t )  ae
k 
k
jk0t
or in the

frequency domain where it is represented by Fourier Series Coefficients (Spectral coefficients).

Reason For Using Exponential Signal As Basic Signal For Building Periodic Signals

The reason why we use the complex exponential signal exp j0t is used as the basic signal to build
the periodic signal x(t) because it is an eigenfunction of an LTI system. An eigenfunction
function like exp j0t when input into an LTI system, the result has the SAME exp j0t but
modified by the eigenvalue at the same frequency.

If the input x(t) into an LTI system is exp j0t , then the output
 

 h( ) x(t   )dt   h( ) exp

j0 ( t  )
y (t )  x(t )* h(t )  h(t )* x(t )  dt
 

Since we are integrating w.r.t  , t is a constant.

  

 h( ) exp j0 (t  ) dt  exp j0t  h( ) exp  jo d . The term  h( ) exp
 jo
y (t )  d is a complex
  

value called eigenvalue. Therefore the output y(t) has exactly the SAME frequency 0 and same
form as the input i.e exp j0t multiply by the eigenvalue. The eigenvalue is said to modify the
amplitude of the input eigenfunction but not the frequency. If an signal ak exp jk0t is input into
an LTI system, then the output will be y (t )  ak H ( jk0 ) exp jk0t where


 h( ) exp
 jo
H ( jk0 )  d is a complex value


3
 
Then if a linear combination of complex exponential signals such as x(t )  ae
k 
k
jk0t
is input
 
into the LTI system, then the output will be y (t )  
k 
ak H ( jk0 )e jk0t  be
k 
k
jk0t

Example

 
Given a signal x(t )  1  sin 0t  2 cos 0t  cos  20t   , find its Fourier Series coefficients.
 4

By using the Euler’s formula cos0t 

1
2
 
exp j0t  exp  j0t , sin0t 
1
2j

exp j0t  exp  j0t , 
 
1  j (20t  ) 
   
1 j (20t  )
j0t  j0t j0t  j0t
x(t )  1  exp  exp  exp  exp   exp 4
 exp 4

2j 2 

By rearranging the coefficients and grouping the terms of 0 , 0 , 20 , 20

The Fourier series Coefficient are

a0  1

 1  1  1  1
a1  1    1  j a1  1    1  j
 2j 2  2j 2

 
1 2 1  j( ) 2
1  j  1  j 
j( )
a2  exp 4  a2  exp 4 
2 4 2 4

ak  0 k 2

Plotting the Fourier Coefficients. ak

We plot the Fourier Coefficients in terms of its magnitude and phase

In the above example

Magnitude Phase

a0  1 ak  0

0.5
2
1
a1  12     1.12 a1  tan 1  26.5
2 1

4
 2
1 0.5
a1  1     1.12
2
a1  tan 1  26.5
2 1


1 2 1 1
1  j  
j( )
a2  exp 4  a2  tan 1  45
2 4 2 1


1  j( ) 2 1 1
a2  exp 4  1  j   a2  tan 1  45
2 4 2 1

Note that for a continuous time signal x(t), the frequency spectra are discrete.

The magnitudes of the Fourier Coefficients are shown below:-

The phases of the Fourier Coefficients are shown below:-

In the above example we use Euler’s formula to find the Fourier Series coefficients. Not many
signals are so conveniently calculated in this manner. We need a more general approach.

5
General Formula For Calculating The Fourier Coefficients

Given a periodic signal x(t ) where x(t) = a
k 
k exp jkot . The fundamental frequency of this

periodic signal is 0 .

We multiply x(t ) by exp  jnot and integrate over a period from at t = 0 to t =T0

  
T0 T0

 x(t )exp
 jnot
dt =    ak exp jkot exp  jnot dt
0 0  k  

On the right hand side equation, interchanging the order of summation and integration yields
T0  T0
T0  
 x(t )exp
 jnot
dt =  a  exp k
jkot
exp  jnot
dt =  ak   exp j ( k n )ot dt 
0 k  0 k   0 
T0

 exp
j ( k  n )o t
The term dt can be evaluated easily by using the Euler’s Formula.
0

T0 T0 T0

i.e a k  exp j ( k n )ot dt  a k  cos(k  n)ot dt  ja k  sin(k  n) t dto

0 0 0

Any sine or cosine function when integrated over a full period is always zero. Therefore for
T0 T0

k n ,  cos(k  n) t dt  0 and  sin(k  n) t dt  0

0
o
0
o

T0 T0 T0 T0 T0

For k = n,  cos(k  n) t dt   cos(0) dt   1dt T

0
o
0 0
0 and  sin(k  n)ot dt   sin(0)dt 0
0 0

T0

For k = n, a k  exp j ( k n )ot dt  a n T0

0

T0

 x(t )exp
 jnot
dt = a nT0
0

T0
1
 x(t ) exp
 jno t
an  dt
T0 0

6
This is the desired relationship between x(t) and an


x(t )  ae
k 
k
jk0t
Synthesis Equation

The synthesis equation tells us how to build a periodic signal x(t) using a linear combination of
complex exponential harmonic related signals.

T0
1
 x(t ) exp
 jko t
ak  dt Analysis Equation
T0 0

The analysis equation tell us how much (how strong) a periodic signal x(t) has at the particular
kth harmonic frequency. (This is similar to the dot product in vector analysis, i.e if you want to
find out how much a vector is in a certain direction, you dot product the vector with that
direction. By doing this you are finding how much is the component of the vector in that
direction)
T0
1
 x(t ) exp
 jko t
The DC value of the signal x(t) is given by substituting k = 0 into a k  dt
T0 0

T0 T0
1 1
 x(t ) exp dt   x(t )dt . The is just the average value of x(t) over a
0
The DC value is a 0 
T0 0
T0 0

period.

Example

Given a periodic square wave as shown and defined over one period

1, t  T1

x(t )   T
0, T1  t 
 2

7
 T T
Because of the symmetry of x(t) about t=0, it is convenient to choose   t  as the interval
2 2
of integration

First we find a0
T
1 1 2T
. a0  
T T1
dt  1
T

T T1
1 1 1
For k  0 we obtain a k   exp  jkot dt   exp  jk0t
T T1 jk0T T1

2 exp jk0T1  exp  jk0T1 

ak   
k0T  2j 

 exp jk0T1  exp  jk0T1 

By Euler’s formula    sin k0T1
 2j 

2
Therefore a k  sin koT1
k0T

2 2 sin koT1
Since T  ak  sin koT1  k 0
0 k 2 k

Plotting The Fourier Coefficients

sin koT1
In the above example, T1 is a constant and the Fourier Coefficients a k  . Let’s say for a
k
specific case, the period T = 4 T1 then the Fourier Coefficients can then be calculated as follow :-

2T1 1
a0  
4T1 2

Note that the square wave given above is an EVEN function. For even function ak=a-k

sin oT1 T 2  2
a1  , Since T1    where the period T 
 4 40 20 0


Substituting T1 
20

8
 
sin
sin oT1 2 1
a1  
  

1
a 1  because the square wave given is an even function.


sin 2oT1
a2 
2


Substituting T1 
20

sin 
a2  0
2

a 2  0 because the square wave given is an even function.

sin 3oT1
a3  ,
k


Substituting T1 
20

3
sin 
sin 3oT1 2  1
a3  
3 3 3

1
a 3   because the square wave given is an even function.
3

sin 4oT1
a4 
4


Substituting T1 
20

sin 2 sin 5oT1

a4  0 a5 
4 5

9
 
Substituting T1 
20

5
sin 1
a5  2  1 a 5 
5 5 5

The magnitude of the ak coefficients of the square wave are plotted below:-

Properties of Continuous Time Fourier Series

We note that for a CT periodic signal x(t), its Fourier Series Coefficients are at discrete
harmonically related frequencies.

When a signal x(t) is real and even then the coefficients a k  ak .Likewise when a signal x(t) is
odd, then a k  ak .

When a signal x(t) is real (i.e do not have any imaginary terms), then ak *  a k

There are also many other properties that can help us when evaluating the Fourier Series
Coefficients.

Linearity

If a signal x(t) has Fourier Series coefficients ak and another signal y(t) has coefficients bk and if
x(t) and y(t) has the same fundamental period, T then

z(t)=Ax(t)+By(t) , then the resultant Fourier Series coefficients of z(t) will be = Aak + Bbk

10
Time Shifting

When a time shift is applied to a periodic signal x(t), the period T of the signal remains
unchanged.

If x(t) has coefficients ak, and if y(t) = x(t-t0) has coefficients bk

 2 
 jk  t0
 jk0t0
Then bk  exp ak  exp  T 
ak

Take note that a periodic signal is shifted in time, the magnitudes of its Fourier Series
 2 
 jk  t0
coefficients remain unchanged as exp  T 
 1 . i.e bk  ak

Time Reversal

The period T of a periodic signal x(t) remains unchanged when the signal undergoes time
reversal.

If x(t) has coefficients ak, and if y(t) = x(-t) has coefficients bk

Then bk  a k

Time Scaling

Time scaling is an operation that in general changes the period of the periodic signal. If x(t) is a
T
periodic signal then x( t ) is periodic with period , and therefore the frequency of each

component will be increased to 0 . Since the time scaling is applied directly to each harmonic
components of x(t), the coefficients remains unchanged as only the frequency has changed.

If x(t) has coefficients ak, and if y (t )  x( t ) has coefficients bk , then bk = ak

However because the fundamental frequency has changed, the Fourier series representation will
be different.

Multiplication

If x(t) has coefficients ak, and if y(t) has coefficients bk


Then z(t) = x(t)y(t) has coefficients hk   ab
l 
l k l

Conjugation and Conjugate Symmetry

11
If x(t) has coefficients ak then x(t)* will have coefficients a* k

If x(t) is real i.e x(t) = x(t)* , then ak  a* k or taking the conjugate of both sides ak*  a k

Application to useful signal:

x(t) real and even

If x(t) is real then ak*  a k ,and if x(t) is also even then ak  a k . If x(t) is both real and even, then
i.e ak  a k  a* .This means that ak must be real and even

x(t) real and odd

If x(t) is real then ak  a* k and if x(t) is also odd, ak  a k . If x(t) is both real and odd, then ak

must be imaginary and odd

Example

Given a square wave with a fundamental period of T= 4 as shown, calculate its Fourier Series
coefficients.

This example is similar to the square wave done earlier. One can easily see that in comparison to
1
the earlier example g (t )  x(t  1)  . We now solve this problem using the alternative way by
2
applying the properties of Fourier series.

12
 2 2 
Based on the time shift property bk  ak exp  jk0t0 . Here t0  1 and 0   
T 4 2

 jk
Therefore bk  ak exp 2

1
To tackle the DC term, we need to focus on the b0. In this case b0  a0 
2

1
Since a0  in the earlier example, b0  0
2

k k
sin sin 
Since ak  2 in the earlier example, b  2 exp  jk 2
k k
k

Therefore

k
sin 
bk  2 exp  jk 2 , k  0
k

0 , k=0

Parseval’ Theorem

The average power over one period for one frequency kω0 :

1 1
 ak exp jk0t dt   ak dt  ak
2
Ptotal 
2 2

TT TT

With a signal x(t) represented in Fourier Series


x(t )  a
k 
k exp jkot

The total average power over a period ;

 2  2
1 1
 ak exp jk0t  
2
Ptotal 
TT x(t ) dt  
TT k  k 
ak

Parseval’s Relation states that

The total average power in a periodic signal equals to the sum of the average powers in all of its
harmonic components.

13
Discrete Time Fourier Series

A discrete time signal x[n] is periodic with period N if

x[n] = x[n+N]

The fundamental period of a discrete time period signal x[n] is N0. This is the smallest integer N
for which the above equation holds

2
The fundamental frequency of a discrete time period signal x[n] is 0 where 0  .
N0
2
Frequently, we simply write 0  treating N as No i.e treating N as the fundamental period
N

Representation of discrete time signal in terms of complex exponential signals

As in the continuous time case, we can also represent a discrete time signal x[n] as a linear
combination of harmonically related complex exponential periodic signal of the form

x[n]   ak exp jk0 n .

k

2
2
, x[n]   ak exp jk0n   ak exp N
jk n
Since 0 
N k k

There are some important points you need to note when dealing with discrete time periodic
signal.

A discrete time complex exponential which differ in frequency by a multiple of 2 are identical.

Given a periodic time periodic signal x[n] and another periodic discrete time signal y[n] which is
2 higher in frequency, x[n] is equal to y[n]

Proof

x[n]  exp jn

y[n] = exp j (  2 ) n  exp jn exp j 2 n  exp jn  x[n] since exp j 2 n  1

14
 2
Since x[n]   ak exp
jk n
N
and any signal which is higher in frequency by 2 are identical, there
k

is no point in performing the summation for k from -  to + . As there are only N number of
distinct discrete frequencies, the summation for k is done over a range of N successive integers
only. This is indicated by expressing the limits of the summation as k = <N> i.e
2 2
x[n]   ak exp 
jk n jk n
N
 ak exp N

k k  N 

where k could on values from 0,1,2,…N-1 or k = 2,3,4,5,,N+1

Summary
2


jk n
x[n]  ak exp N
Synthesis equation (to build the signal using complex exponential)
k  N 

1
ak 
N

n  N 
x[n]exp  jk0 n Analysis equation (find the individual spectral coefficient)

Notice that in the analysis equation, the index for the summation is n and not k

The Fourier Series coefficients ak are referred to as the spectral coefficients of x[n]. These
coefficients specify the decomposition of x[n] into a of N harmonically related complex
exponential periodic signals.

1
Since k in ak  
N n  N 
x[n]exp  jk0 n varies only over a range of N successive integer, we can

conclude that the values of ak repeat periodically with period N.

Since there are only N distinct complex exponentials that are periodic with period N, the
discrete-time exponential Fourier Series representation is a finite series with only N terms. In
comparison with the continuous time case, the Fourier series is an infinite series.

15
Properties of Discrete Time Fourier Series

Property Periodic Signal Fourier Series Coefficients

Linearity Ax[n]+By[n] Aak+Bbk

 2 
 jk   n0
Time shifting x[n  n0 ] ak exp  N 

Conjugation x*[n] a* k

Time Reversal x[ n] a k

n ak
Time Scaling x   if n is a multiple of m
m m

0 if n is not a multiple of m

Multiplication x[n]y[n]  ab
l  N 
l k l

 
 2 
 jk 
First difference x[n]-x[n-1] 1  exp  N   ak
 
 

Real and Even Signals x[n] real and even ak real and even

Real and Odd Signals x[n] real and odd ak purely imaginary and odd

Parseval’s Relation For Discrete Time Periodic Signals

1
 
2 2
x[n]  ak where N is the period and ak are the Fourier Series coefficients.
N n  N  k  N 

Parseval’s relation states that the average power in a periodic signal equals to the sum of the
average powers in all of its harmonic components. For discrete time there are only N distinct
harmonic components, the right hand side can be taken over any N consecutive values of k

16
Example 1

 2   2   4 
Given a discrete time signal x[n]  1  sin   n  3cos   n  cos  n 
 N   N   N 2

Calculate its Fourier Series coefficients.

Firstly we take note that the fundamental period of the signal is N

2
The fundamental frequency is 0 
N

We then express x[n] in terms of the complex exponential using the fundamental frequency

1 n  
2 2 2 2  4    4  
1  j n j n 3 j n j n j n   j
x[n]  1   exp N
 exp N
   exp N
 exp N
   exp  N 2
 exp  N 2

2j   2   2  

1 n 
2 2 2 2   2    2 
1  j n j n 3 j n j n j j 2 n j  j 2
x[n]  1  exp
N
 exp N
  exp
N
 exp N
  exp exp
2  N 
 exp exp  N  
2
2j   2  2  

Collecting terms

2 2     2  2
3 1  j n 33 1  j n 1 j  j 2 n  1 j   j 2 n
x[n]  1     exp N
    exp N
  exp  exp
2  N 
  exp  exp
2  N 

2 2j 22 2j  2  2 

Thus the Fourier Series coefficients are

a0  1

3 1 3 1
a1     j
2 2j 2 2

3 1 3 1
a1     j
2 2j 2 2

1
a2  j
2

1
a2  - j
2

17
The rest of the aks are zero in the interval of summation in the synthesis equation. Remember that
the Fourier Series coefficients are periodic with period N. Also note that in this example ak  a* k
This reveals that the discrete time periodic signal x[n] is a real signal.

Plotting The Fourier series Coefficients

The Fourier Series coefficients are usually plotted in terms of their magnitude a 2  b2 and
b
phase tan 1 .
a

Fourier Series coefficients Magnitude Phase

a0  1 1 0

3 1 3 1 10
a1     j 18.4
2 2j 2 2 2

3 1 3 1 10
a1     j 18.4
2 2j 2 2 2

1 1
a2  j 90
2 2

1 1
a2  - j 90
2 2

Magnitude Plot

Phase Plot

18
Example 2

Calculate the Fourier Series coefficient of the square discrete time periodic signal as shown.

N1 N1 2
1 1  jk
 x[n]exp  jk0n   x[n]exp
n
ak  N
N  N1 N  N1

N1 2 N1 N1
1  jk 1 1
 x[n]exp  x[n]  1
n
For k = 0,  N,  2N, substitute k = 0 ak  N

, N  N1 N  N1 N  N1

2 N1  1
ak  for k = 0,  N,  2N
N

To calculate ak for k  0,  N,  2N , it is convenient to choose the period N to include –N1 to

N1
N1 N1 2
1 1  jk
 x[n]exp  jk0n   x[n]exp
n
ak  N
N  N1 N  N1

Let m = n+N1
2 N1 2
1  jk  m  N1 
ak 
N
 exp
m0
N

jk 2  N1

N
Factoring out by taking out the constant term exp (term not involving m ) out of the
summation
jk 2  N1

2 N1 2 m
exp N  jk
ak 
N
 exp
m0
N

1   N 1 N
Note the summation of a standard series    n

n 0 1

19
 jk 2  N1
  jk
2 (2 N1 1)

exp N  1  exp N

ak 
N   2 
 jk  


 1  exp
 N 

1
2 ( N1  )
 jk 2
N
Factoring out the term exp in the numerator

jk 2  N1 jk 2  N1
1
2 ( N1  )  1
2 ( N1  )
1
2 ( N1  ) 
  jk
2 (2 N1 1)
  jk 2
 exp jk 2  jk 2

1 exp N
 1  exp N
 1 exp N
.exp N

N
 exp N

ak    2   N  2 
N  1  exp  jk  N     jk   
   1  exp  N 

 

 2 
 jk  
 2N 
Factoring out the term exp in the denominator

 
jk 2  N1  
 2 
1
jk 2   N1  
 1
 2
 
 jk    2

2N 

N
N
exp  exp exp 
 
ak 
1  
2N   2N  
 2   2   2 
N  jk   jk 
2N 
  jk  

exp exp  exp 

 

  1 
 2 k  N1  2  
sin   
 N 
1 
 
ak  ak for k  0,  N,  2N
N k 
sin  
 N 

20
Continuous Time Fourier Series and LTI Systems

The complex exponential signal exp j0t is used as the basic signal to build the periodic signal x(t)
because it is an eigenfunction of an LTI system. An eigenfunction function like exp j0t when
input into an LTI system, generates an output of the SAME exp j0t but modified by the
eigenvalue at the same frequency.

If the input x(t) into an LTI system is exp j0t , then the output
 

 h( ) x(t   )dt   h( ) exp

j0 ( t  )
y (t )  x(t )* h(t )  h(t )* x(t )  dt
 

Since we are integrating w.r.t  , t is a constant.

  

 h( ) exp  h( ) exp  h( ) exp

j0 ( t  ) j0t  jo  jo
y (t )  dt  exp d . The term d is a complex
  

value called eigenvalue. Therefore the output y(t) has exactly the SAME frequency 0 and same
form as the input i.e exp j0t multiply by the eigenvalue. The eigenvalue is said to modify the
amplitude of the input eigenfunction but not the frequency. If a signal x(t )  ak exp jk0t is input
into an LTI system, then the output will be y (t )  ak H ( jk0 ) exp jk0t where


 h( ) exp
 jko
H ( jk0 )  d is a complex value


Example For Illustrating The Role of Eigenvalue Of LTI System

Suppose that the input to an LTI system is x(t )  10exp j 3 t . The angular frequency of this signal,
 is 3 . Let’s say the value of the eigenvalue of the LTI system at this particular frequency is
2- j2, we can write then write H ( j 3 ) = 2- j2. The eigenvalue of the the LTI system at this
particular frequency written as H ( j 3 ) modifies the input signal x(t) at the frequency  of 3
and produce the output signal y(t).
2 j
tan 1 
Converting H ( j 3 ) into the polar form i.e 2  2 j  22  22 exp 2
 2 2 exp 4

Therefore the output y(t) is simply the eigenvalue of the system at this particular frequency of
3 multiply by the input frequency x(t) i.e. y (t )  H ( j3 ) x(t )
j 
 j (3 t  )
j 3 t
Therefore y(t )  2 2 exp 4
.10exp  20 2 exp 4

21
Frequency Response of an LTI System

When the input x(t) is a complex exponential signal whose frequency is  , the output is also a
complex exponential of having EXACTLY the same frequency. However the effect of the LTI
system on a complex exponential signal depends on the frequency of the input x(t) i.e the
eigenvalue of H ( j ) of the system is not a constant but a function of the input signal frequency
 . We will get a different value of H ( j ) if we apply an input x(t) with a different  each time.
If we let the frequency  vary, we can define the frequency response of a continuous-time LTI
system as

H  j    h  t  exp
 jt
dt


In general we want a functional form of H  j  so that we can plot a graph of H  j  as a

function of  . By doing this, we are then able to see the effect of the system to any complex
exponential input.

SUMMARY: Knowing the LTI’s system’s frequency response H ( j ) is important as it allows

us to calculate the output signal easily given any input complex exponential signal. Also we have
learnt that the impulse response h(t) (a function of time ) is very closely related to the frequency
response H ( j ) (a function of frequency).

Example

Given an LTI system with an impulse response h(t )  2exp 2t u (t ) calculate its frequency
response H  j  .


H  j    2 exp
2 t
u (t ) exp  jt dt


Because of u(t) in the expression for h(t), the integration is between 0 and  only

 
2exp2t  jt 2 2
H  j    2exp 2t
exp  jt
dt   0 
0
2  j 0
2  j 2  j

Note that the frequency response of the LTI system H  j  is a function of frequency 

22
Plotting The Frequency Response H  j 

The frequency response H  j  is a complex function that depends on  . As H  j  is

complex, it is best to plot it in two graphs i.e the magnitude plot vs frequency H  j  and the
2
phase vs frequency H ( j ) .Given an LTI system with a frequency response H  j   ,
2  j
plot its magnitude and phase response.

2 2  j 4  j 2
H  j   . 
2  j 2  j 4  2

16  4 2 4(4   2 ) 2
H  j    
4 2
4 2
4  2

2 2
Short cut: H  j    can be calculated directly by taking the magnitude of
2  j 4  2
the denominator.

 
H ( j )  0  {2  j}   tan 1  
2

The range of the frequency axis is  <  <  since the continuous-time complex exponential
signals can have any frequency in that range. If we evaluate a few crucial frequencies such as
  0, 2, 20,  we obtain the values shown below. The value of   2 is chosen by looking at the
2
denominator of . At the value of   2 , the magnitude H  j  would have dropped
4  2
to 0.707 compared to the value of H  j  at   0

 H  j 

0 1

2 1
2   0.707
8 2

2
20  0.0995
404

 0

23
The figure below shows a complete plot of the magnitude and phase of the frequency response
H  j  over the range 8    8 . By looking at the magnitude response, one can conclude
2
that the LTI system whose frequency response H  j   is a Low Pass Filter
2  j

Response to Periodic Input Signals

When the input signal x(t) into an LTI system is periodic, it is easy to see that the output signal
of the system will also be periodic.

Then if a linear combination of complex exponential signals such as x(t )  ae
k 
k
jk0t
is input
 
into the LTI system, then the output will be y (t )   a H ( jk )e
k 
k 0
jk0t
 be
k 
k
jk0t

where bk  ak H ( jk0 )

Thus y(t) is also periodic with the same fundamental frequency as x(t). If ak  is the set of
Fourier series coefficients for the input x(t), then ak H ( jk0 ) is the set of coefficients for the
output. The effect of the LTI system is to modify individually each of the Fourier series
coefficients of the input through multiplication by the value of the frequency response at the
corresponding frequency.

24
Example : Given an LTI system with impulse response h(t )  expt u (t ) and an input
x(t )  1  cos t  cos8t find the output y(t)

First we find the Fourier series coefficients of x(t). The first thing to do is to calculate the
fundamental frequency 0 =1. [ 1 for cos t = 1 and 2 for cos8t  8. Overall 0  1]

Step 1: Find the Fourier coefficients for the input signal x(t)

x(t )  1  cos t  cos 8t  1 

1
2
 exp t  exp  t    exp8t  exp 8t 
1
2

1 1 1 1
Therefore c0  1, c1  , c1  , c8  , c8 
2 2 2 2

Step 2 : Find the frequency response H  jk0 

 
H  jk0    h(t ) exp  jk0t dt   exp  t exp  jk0t dt
 0


exp  0
  1 jk t
1 1
H  jk0    exp  0  1 
 1 jk0 t
dt  =
0
 1  jk0  0 1  jk0 1  jk0

If you try to sketch the frequency response, you will find that this is a Low Pass Filter.

Step 3:

Find the output y(t)


y (t )   a H ( jk )e
k 
k 0
jk0 t

1 1  1 1   jt 1 1  1 1  8 jt
y (t )  1    exp  
jt
 exp    exp  
8 jt
 exp
2  1 j  2  1 j  2  1 8 j  2 1 8 j 

By looking at the Fourier series coefficients of the output y(t), the LTI system is clearly a Low
Pass Filter.

1 1   j 0.79 1 1 
k  1 H(j)=   =0.35exp k  1 H(-j)=   =0.35exp
j 0.79

2  1 j  2  1 j 

1 1   j1.45 1 1 
k 8 H(j8)=   = 0.06exp k  8 H(-j8)=   =0.02exp
j1.45

2  1 8 j  2  1 8 j 

25
Discrete Time Fourier Series and LTI Systems

In a discrete time LTI system with an impulse response h[n], when an input discrete time signal
x[n] is applied, one can compute the output y[n] in a similar manner to the case for the
continuous time Fourier Series.

For a discrete time periodic signal of one frequency 0 as the input x[n]  exp j0n ,

then the output will also be a periodic signal with one frequency y[n]  H (exp j0 ) exp j0n

If we now have an input x[n] which is periodic which consists of a linear combination N of
frequencies
 2 
jk  n
x[n]  
k  N 
ak exp jk0 n  
k  N 
ak exp  N 

then

 j k 
 2   2 
jk  n
y[n]   ak H 

exp  N 


exp  N 

k  N   

 2
jk 

n  j k 
 2 

y[n]   bk exp  N 
where bk  ak H  exp  N  
 
k  N   

Note that the frequency response for the discrete time case is written as

H [exp j ]   h[n]exp
n 
 j n
where h[n] is the impulse response. Note that the summation for n is

from  to 

Example

 2 n 
n
1
Given an LTI system with impulse response h[ n]    u[ n] and an input x ( n)  cos  ,
2  3 
find the output y(n)

Step 1: First we find the Fourier series coefficients of x(n). The first thing to do is to find out the
2
fundamental frequency 0 . For this example, 0  .
3

 2 n  1  
 2 n    2 n 
j  j
x(n)  cos     exp  3 
 exp  3 

 3  2  


26
The Fourier coefficients are

1 1
a0  0, a1  , a1 
2 2

Step 2 : Find the frequency response H  e j 

 n  n
1 1 
H  e      exp  j n    exp  j 
j

n 0  2  n 0  2 

N
1   N 1 
1   N 1 1
Note that  n 
n 0 1
, When N     n 
n 0 1

1
. The term  N 1  0 applies

only when   1

1
In our case here ,     exp  j  1 as exp  j  1
2

 n
1 
H  e     exp  j  
j 1
n 0  2  1  1 exp  j
2

Step 3:

Find the output y(n)


y[n]   a H (e  )e
k 
k
j jk0 n

   
1  1 
 2 n   2 n 
1 j  1  j 
y[n]   2
 exp  3    2
 exp  3 
2 1 j
 2 1 j

 1  exp 3   1  exp 3 
 2   2 

27
Systems Characterised By Differential Equation and Difference Equations

We now apply what we have learned so far to calculate the frequency response of second order
differential systems and difference equations and also their outputs given input excitations of
periodic signals.

Example 1

Given an LTI system described by a second order differential equation

2
d y (t ) dy (t )
  y (t )  x(t ) and the input voltage is x(t )  sin t . Calculate the system’s output
dt 2 dt
y(t).

Step 1. Find the frequency response of the system

We purposely apply an eigenfunction input x(t )  exp jt into the system and the output is simply
y(t )  H  j  exp jt

Substituting the output y(t) into the differential equation, we obtained

 2 H ( j)exp jt  j H  j  exp jt  H  j  exp jt  exp jt

Eliminating exp jt , we obtain  2 H ( j )  j H  j   H  j   1

1
 H ( j )  . This is the frequency response of the system.
  j  1
2

Step 2: Find the Fourier Series coefficients of the input x(t).

Since the input x(t )  sin t , its fundamental frequency 0  1 . We also know that sin t can be

decomposed into 2 frequencies based on Euler’s formula i.e sin t 

1
2j
 exp jt  exp  jt 

1 1
a1  a 1  
2j 2j

Then y (t )  a1 H ( j ) exp jt  a1H ( j ) exp  jt

1 1 1 1 1
y (t )  exp jt  j exp  jt   exp jt  exp  jt   cos t
2j j 2j 2 2

28
Example 2

Consider a causal discrete time LTI system whose input x[n] and output y[n] are related by the
difference equation

1
y[n]  y[n  1]  x[ n]
4

Find the Fourier Series representation of the output y[n] when the input
   
x[n]  cos  n   2 cos  n 
4  2 

Step 1. Find the frequency response of the system

We purposely apply an eigenfunction input x(n)  exp jn into the system and the output is simply
y (n)  H  e j  exp j n

Substituting the output y(n) into the difference equation, we obtained

H (e j ) exp j n  H  e j  exp    exp j n

1 j n 1

Eliminating exp jn , we obtain H (e j )  H  e j  exp  j  1

1
4

1
 H (e j )  . This is the frequency response of the system.
1  j
1    exp
4

Step 2: Find the Fourier Series coefficients of the input x(n).

   
Since the input x  n   cos  n   2 cos  n 
4  2 


The fundamental frequency 0  . (The fundamental frequency in this case was found from
4
visual inspection. If the 2 frequencies are harmonically related to each other, and we simply take
smallest frequency).Also for discrete time case, we need to calculate the fundamental period.

0  1 k
   . The fundamental period N=8
2 4.2 8 N

29
By apply Euler’s formula,
   
1 j n   j n 
x  n   exp 4  exp 4    exp 2  exp 2  .
j n j n

2   
   
1 1 j n j n
x  n   exp  exp
j n j n
4 4
 exp  exp 2
2
2 2

1 1
a1  a 1 
2 2

a2  1 a 2  1

Step 3. Calculate the output y[n]

 2 

 
jk  n
y[n]  y[n]  
k  N 
bk exp  N 
where b k  ak H exp jk0

   
1  1  1 1 
b1  a1 H (e j0 ) =     
2  1  1 exp  j0  2  1  1 exp  j 4 
 4  
 4 

   
 j0 1 1  1 1 
b1  a1 H (e ) =     
2  1  1 exp j0  2  1  1 exp j 4 
 4  
 4 

   
 1   1 
b2  a2 H (e j 20 ) =    
1
 1  exp  j 20   1  1 exp  j 2 

 4   4 

   
 j 20
 1   1 
b2  a2 H (e ) =   
1
 1  exp j 20   1  1 exp j 2 

 4   4 

       
j  n  j  n j  n  j  n
4 4
exp exp exp  2  exp  2 
Therefore y[n]  j
 j
 j
 j
 1    1  1  1
2 1    exp 4  2 1    exp  1   4  exp 1    exp 2
4 2

 4    4     4

30