Chapter

DISCRETE TIME
FOURIER SERIES 6
6.1 Introduction

In the previous chapters we have analyzed frequency of continuous time signals from now we will start
analyzing the frequency domain of discrete signals. The discrete signals are obtained by sampling the continuous
the signal

Continuous Discrete
Sampling
Time signal Time signal

Figure 6.1 Sampling of continuous time signal

In chapter 5 we have seen that when sampling is done that is x(t) is multiplied by impulse train then the
frequency domain become periodic with period s. That is sampled signal will always have periodic frequency
domain. Discrete time fourier series and discrete time fourier transform are used for frequency analysis of
perodic and non-perodic discrete signal. In both the transforms the frequency domain is periodic. In we take
fourier transform of x(t) as X() and sampled signal xp(t) and it is fourier transfom as Xp(). If sampling
frequency is s then Xp() will be periodic with period s.

Xp()
X()
A
Sampling

B 0 B  s B B 

0 s

Figure 6.2 Frequency analysis of sampling process.

 is the continuous frequency and  is discrete frequency and to convert  to  we use scaling.
s  2 ... (1)
That is s of continuous frequency will become 2 in discrete frequency.
Thus s  2

GATE MASTERS PUBLICATION

 270 Signals & Systems
2
Or TS  2

2
For any    .
s

That is   TS  ... (2)

Thus  in continous frequency domain can be changed to dicrete frequency whose value will be TS.

REMEMBER  of continuous frequency domain can be change to dicrete frequency domain by simply multiplying
by TS.
 = TS
And to convert discrete frequency to continuous frequency can be done by division of TS.

6.2 Discrete Time Fourier Series(DTFS)

Using DTFS any discrete time periodic signal with period N can be represented as a linear combination of
N orthogonal peridic complex exponential signals. The fundmental frquency is wo = 2/N. The complex
exponential function is
jk  n
k ( n ) = e o ... (3)
2
jk n
N
= e ... (4)
From equation (4) we can see that
0(n) =N(n)
That is the function is same for k = 0 and k = N, similarly function will be same for k = 1 and k = N + 1.
Thus we can see that k(n) function will be distinct only when k is between O to N1. Only N distnict function
can be generated for k(n).

Study Note
A very important difference is seen between discrete and continuous complex exponential function. In case of
jk  t
continuous complex exponential function e o we have infinite functions for  < k <  but discrete complex
jk  n
function e o will give only N distinct function.

Thus using DTFS we can express any periodic signal x[n] periodic with period N in terms of linear combination
of N complex exponentials

x[n] =  Xk k ( n) ... (5)

k  N 

 x(n) =  X k e jk o n ... (6)

k N 

We can see in above equation (5) and (6) that summation is over N, thus DTFS coefficients are finite and
only N terms exist in DTFS of a periodic signal with period N.

Study Note
 k(n) = k+N(n) that is complex exponential function is periodic with period N that is function is same
for values k and k+N.
 X0 is coefficient of 0(n) and XN is coefficient of N(n) thus XN = X0. similarly Xk is coefficient of k(n) and
Xk+N is coefficient of k+N(n) thus Xk+N = Xk. Thus Xk is also periodic with period N.

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 271

 The DTFS of dicrete periodic signal of period N generate coefficients Xk which are also discrete and
periodic with period N.
To evaluvate DTFS coefficients the formulae is

  jo kn
Xk =
N
 x  ne
n  N 
... (7)

Study Note
There are no dicrichlet condition for DTFS, the only condition is that signal should be periodic. The summation
will always give finite result.

6.2.1 Magnitude and phase spectrum of discrete time periodic signal

The fourier series consists of N components
X0 , X1 e j o n , X2 e j 2o n ,...,XN-1 e j ( N 1)o n
The frequencies of these components are 0, o, 2o,...,(N1)o where o = 2/N. The amount of the kth
harmonic is Xk (the fourier coefficient). We can plot this amount Xk as a function of index k or . Such a plot is
called the fourier spectrum of x(n). In general, the fourier coefficients Xk, are complex and be represnted in the
polar form as

Xk = X k e j  X k
The plot of |Xk| versus  is called the magnitude spectrum and that of Xk versus  is called the angle (or
phase) spectrum. These two plots together are the frequency spectra of x(n). Knowing these spectra, we can
reconstruct or synthesize x(n) according Eq.(6).

Study Note
The results are very similar to the representation of a continuous-time periodic signal by an exponential fourier
series except that, the bandwidth of the continuous-time perodic signal is infinite and consists of an infinite
number of exponential components (harmonics). The spectrum of the discrete-time periodic signal, in contrast, is
bandlimited and has at most N components.

6.2.2 Finding DTFS coefficients of a signal

Generally we can use Eq.(7) to find Xk of any given signal x[n], but if x[n] has sinusoidal or complex
exponentials then we cannot use formulae to find Xk. To find Xk in such cases we will follow these steps
Steps 1. Find the period of the signal, if the signal is non periodic then we cannot find DTFS coefficients.
If it is periodic then move to step 2
Steps 2. Let the signal is periodic with period N. Now express all the sine and cosine signals in terms of
complex exponentials

e j  e j e j  e j
sin = cos=
2j 2
jk o n
Steps 3. Now compare each term with Xk e and find value of k and Xk. By comparing each term we
can easily get Xk.

GATE MASTERS PUBLICATION

272 Signals & Systems

Example 6.1

Determine the fourier series coefficients of the signal x ( n ) and plot its magnitude and phase
spectrum.

 2π   2π   4π π
x ( n ) = 1+sin  n +3cos  n +cos  b+ 
 N   N   N 2
Solution 6.1

The given signal is periodic with period N and frequency o = 2/N and it can be written as

 
x(n) = 1  sin  o n   3cos  o n   cos  2o n  
2  

 e j o n  e  j o n   e j o n  e  j o n   e j 2o n/2  e  j 2o n/2 

x(n) = 1+    3    
 2j   2   2 

3 1  j o n 3 1 
 j o n 1  1 
= 1   2  2 j e   e   e j  /2  e j 2o n   e  j  /2  e  j 2o n
  2 2j  2  2 
Comparing the above equation with.
x(n) = X 0 + X 1e j o n + X 1e  j o n + X 2e j2 o n  X -2e  j2 o n
We get
X0 = 1

3 1 3 1 10
X1 =    j    tan1 1
2 2j 2 2 2 3

3 1 3 1 10
X1 = 2  2 j  2  2 j  2  tan 13
1

1 1
X2 = j  90
2 2

1 1
X2 =  j    90
2 2
The magnitude spectrum and the phase spectrum are depcted in figure (a) and (b), respectively.

|Xk|

10 / 2

1
0.5

N 11 0 1 2 N k

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 273

Xk

/2

N 10 1 2 N k

/2

Figure (a) Magnitude spectrum, (b) Phase spectrum

Example 6.2

Evaluate the fourier series for the discrete-time periodic square wave x ( n ) shown in figure

x(n)

N N1 0 N1 N n

Figure Discrete-time periodic square wave

Solution 6.2
The signal is periodic with period N

1  jo kn
 Xk =
N
 x  ne
n  N 

Here o = 2/N
N1 
1 j . kn
 Xk = N  1.e
 N1
N

  j 2N k (2 N1 1) 
 e  1
1  j N .kN1  
= .e
N   j 2N k 
e  1
 

GATE MASTERS PUBLICATION

274 Signals & Systems

2  N1 1 
j 2 k
N  2  j .
Taking e common from numerator and e N 2 common from denomination

  j 2N k  N121   j 2N k  N121  

   
1 e e 
 Xk =  2  k 2  k 
N j . j .
 e N 2
 e N 2 
 

  2  N1  1   
 sin  N .k  2   
1    
 Xk   , 0  k  N 1
N  2 k  
 sin  .  
  N 2 

Example 6.3

Find the DTFS coeficients of the N-periodic impulse train



x(n) =    n- mN 
m=

As shown in figure

xn)

N 0 N 2N n

A discrete-time impulse train with time period N.

Solution 6.3
The signal is periodic with period N and frequency 0 = 2/N. The given signal for one period may be
wrietten as

1, n  0
x(n) = 
0,1  n  N  1
= (n)
Therefore, the fourier series coefficients are
N 1
1  jk o n
Xk =
N
 x  n e
n 0

N 1
1  jk 2  / Nn 1   jk  2  / N n  1
=
N
   n e
n 0

N  e  n0

N

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 275

Example 6.4

Find the frequency domain representation of the signal depicted in figure

xn)


½

6 5 0 1 2 3 45 6 78
n
1/2

Solution 6.4

2
The signal has period N = 5, so o = . Also, the has odd symmetery, so we sum over n = – 2 to
5
n = 2. Thus

1 1 2
Xk =  x  n e jko n =  x  n e  jk 2  /5 n
N n  N  5 n  2

jk 4  /5
1  x  2  e  x  1 e jk 2  /5 
=  
5   x  0   x 1 e  jk 2  /5  x(2)e  jk4  / 5 

Using the values of x(n), we get

1  1 jk 2  /5 1  jk 2  /5  1   2  
Xk =  e  e   1  j sin  k 
5 2 2  5  5 
The one period of the DTFS coefficients Xk, k = -2 to k = 2 are

1 sin  4 / 5
X-2 = j  0.232e j 0.531
5 5

1 sin  2 / 5
X-1 = j  0.276e j 0.760
5 5

1
X0 =  0.2e j 0
5

1 sin  2 / 5 
X1 = j  0.276 e j 0.760
5 5

1 sin  4 / 5 
X2 = j  0.232 e j 0.531
5 5
Example 6.5

Determine the time-domain signal x ( n ) form the DTFS coefficients depicted in figure

GATE MASTERS PUBLICATION

 276 Signals & Systems

Xk
 Xk|
 

12 3 4 5
7 6 5 3 6 7 8 9 k 7 6 5 4 1 0 1 2 3 4 5 6 7 8 9 10 k
0




Figure (a) Phase spectrum, (b) Magnitude spectrum

Solution 6.5

The DTFS coefficients have period N = 9, so o = 2/9. From figure, the DTFS coefficients over the
interval k =4 to k = 4 are
Xk = | X k | e jX k , 4  k  4

X-4 = 0, X 3  1e j 2  /3 , X 2  2e j  /3 , X 1  0,

X0 = 1ej, = X 1  0, X 2  2e  j  /3 , X 3  1e  j 2  /3 , X 4  0
Thus we apply equation to get x(n)

x(n) =  X k e jko n
k  N 

= X k e jk 2  /9 n
k 4

= X 4 e  j 8  /9 n  X 3e  j 6  /9 n  X 2 e  j 4  /9 n  X 1e  j 2  /9 n

X 0  X 1e j 2  /9 n  X 2 e j 4  /9 n  X 3 e j 6  /9 n  X 4 e j 8  / 9 n
= e j 2/3 e j 6/9n  2e j/3e j 4/9n  e j  2e j/3e j 4/9n  e j 2/3e j 6/9n

 e j  6  /9 n 2  /3  e j  6  /9 n 2 /3   e j  4  /9 n  /3  e j  4  /9 n  /3 

= 2   4    1
 2 2
   

 6 2   4 
x(n) = 2 cos  n   4 cos  n   1
 9 3   9 3

6.3 Properties of DTFS

The properties of DTFS and continuous time fourier series are similar. Let Xk are DTFS coefficients of x[n]
that is
x[n]  Xk

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 277

6.3.1 Linearity
It x(n) and y(n) denote two periodic signals with period N, and
x(n)  Xk y(n)  Yk
Then z(n) = ax(n) + by(n) Zk = aXk + bYk ... (9)
Proof -The fourier series coefficients of z(n) is given by

1 1
Zk =  x n e jkon   ax n  by  n  e jkon
N nN N nN

1 1
= a  x  n e jkon  b  y  ne jkon
N nN 
 N nN 

Xk Yk

= aXl+bYk

6.3.2 Time Shifting

When a time shift is applied to a periodic signal x(n), the period N of the signal is preserved.
If x(n)  Xk
Then y(n) = x(nn0)  Yk = Xk e  jk o n0 ... (10)
One consequence of this property is that, when a signal is shifted in time, the magnitudes of its fourier series
coefficients remain unaltered. That is, |Yk|=|Xk|
Proof - by definition,
N 1
1 1
Yk =  y  n e  jk o n   y n e  jk o n

N n  N  N n0

N 1
1  jk o n
=
N
 x n  n e
n 0
0

A change of variables is perfomed by letting m = ( n  n 0), Which also yields ( m  n 0) as( n  0 ),

and(m(N1n0)) as (n(N1)). Therefore,
N 1 n0
1  jk o ( m+n0 )
Yk =
N
 x  m e
m  n0

N 1 n0
1  jk o m  jk o n0
Yk =
N
 x  m e
m  n0
e

Yk = Xk e  jk o n0

Study Note
We can see that in time shifting the effect is see going with same sign
x[n]  Xk
x[n] Xk
same sign

x n  n0 X k e  jn0o k

GATE MASTERS PUBLICATION

 278 Signals & Systems

Time shifting only change the phase of DTFS coefficients

6.3.3 Frequency shifting

If x(n)  Xk
Then y(n) = e jM o n x(n)  Yk = XkM ... (11)
Proof - by definition,

1  jkon 1 N 1
Yk = 
N nN 
y  n  e   y  ne jkon
N n0
N 1
1 jM o n
= e x  n  e  jk o n
N n 0

N 1
1  j  k  M  o n
=
N
 x n e
n0
 X k M

Hence, a frequency shift corresponds to multiplication in time domain by a complex exponential whose
frequency is equal to the time shift.

Study Note
The frequency shifting property is easy to remember, here the effect goes with opposite sign.
x[n]  Xk
Then

x[n] Xk
apposite sign
 jo Mn
e x (n ) X k M

6.3.4 Time Reversal

If x(n)  Xk
Then y(n) = x(n)  Yk = Xk
In other words, the time reversal applied to a discrete-time signal results in a time reversal of the corre-
sponding sequence of fourier series coefficients.
Proof

1 1 N 1
Yk =  y  n e jk on   y  n  e jk on
N n  N  N n 0

N 1 0
1 1
=  x  n e jkon   x  m  e j   k o m  X  k
N n 0 N m   N 1

An interesting consequence-of the time reversal property is that if x(n) is even then its fourier series coef-
ficients are also even, i.e.,

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 279

if x(n) = x(n)
Then Xk = Xk ... (13)
Similarly, if x(n) is odd, then so are its fourier series confficients, i.e.,
if x(n) = x(n)
Then Xk = Xk ... (14)

6.3.5 Time scaling

Because of the discrete-time nature of the time index for discrete-time signals, the relation between time
scaling and fourier series coefficients in discrete-time takes on a different form its continuous-time counterpart.
If we try to define the signal x(an), we run into difficulties if a is not an integer. Therefore, we cannot slow down
the signal by choosing a < 1. On the other hand, if we let a be an integer other then ±1, we do not merely speed
up the original signal. For example, since n can take on only integer values, the signal x(2n) consists of the even
samples of x(n) alone. Let m be a positive integer, and the signal is defined as

 x  n / m  , if n multiple of m
x(m)(n) =  ... (15)
0, if n is not a multiple of m
xm(n) can be obtained from x(n) by placing (m-1) zeros between successive value of the original signal.
Intuitively, we can think of x(m)(n) as a slowed-down version of x(n). If x(n) is periodic with period N, then
y(n) = x(m)(n) is also periodic with period mN.
Now, if x(n)  Xk
1
Then y(n) = x(m)(n)  Yk  Xk ... (16)
m
The fourier series coefficients Yk = (1/m) Xk are also periodic with period mN.
Proof - the fourier series coefficients of y(n) = x(m)(n) [y(n) is periodic with period mN. and fundamental
frequency o/m] are given by
m N 1
1
Yk =  y (n)e  jk ( o / m ) n
mN n 0

m N 1
1 n  jk ( o / m ) n
=
mN
 x  m e
n 0

A change of variables is performed by letting r = n/m, which also yilds r = 0 as n = 0, and r = N1 as n =
m(N1).

1 1 N 1 1
Therefore, Yk =  x(r )e  jko r  X k
m 
N n 0  m
Xk

6.3.6 Periodic convolution

If x(n)  Xk,
y(n)  Yk

Then z(n) = x(n) y(n) = 

r  N 
x(r ) y (n  r )  Z k  NX k Yk ... (17)

GATE MASTERS PUBLICATION

 280 Signals & Systems

Where  denotes the periodic convolution.

Proof - for periodic signals with the same period, a special form of convolution, known as periodic convolu-
tion, is defined as

z(n) = 
r  N 
x(r ) y (n  r )

It is easy to show that z(n) is periodic with period N and the periodic convolution is comutative and associa-
tive. Thus, we can write z(n) in a fourier series representation with coefficient

1 1  
Zk   z(n)e jkon =    x(r ) y (n  r ) e  jk o n
N nN  N n  N  r  N  

1 
Zk =  x(r )   y (n  r ) e  jko n 
n  N  N r  N  
From the time shifting property, i.e.,
if y(n)  Yk,

Then y(nr) = Yk e  jr o n
We have

  jr  n
Zk = N N 
x(r )e o Yk  NX k Yk

r  N 
Xk

Study Note
Convolution in time domain is multiplication in freqency domain.

6.3.7 Multiplication
If x(n) and y(n) are periodic signals with the same period N, and
x(n)  Xk,
y(n)  Yk,

Then z(n) = x(n)y(n) = Z k  

r  N 
X r Yk  r ... (18)

Zk may be interpreted as the periodic convolution between the two periodic sequences of fourier coefficients
Xk and Yk.
Proof - consider the signal z(n).

z(n) = x(n)y(n) =  X r e jro n  Ym e jmo n

r  N  m  N 

=  Xr  Ym e j ( m  r ) o n
r  N  m  N 

A change of variables is perfomed by letting k = m + r, which also yileds m = (kr), (kr) as (m0), and
[k(r+N1)] as [m(N1)]. Therefore,

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 281

r  N 1

z(n) =  Xr Y k r e jk o n
r  N  k r

=  Xr  Yk  r e jk o n
r  N  k  N 

 
z(n) =    X r Yk  r e jk o n   Z k e jk o n
k  N  r  N   k  N 

Thus, Zk = 
r  N 
X r Yk r

Study Note
Multiplication in time domain is convolution in frequency domain.

6.3.8 First differece

If x[n] is periodic signal with period N and it is DTFS coeffients are Xk. A new signal y[n] is related to x[n]
y[n] = x[n]x[n1] ... (19)
Signal g[n] will also be periodic with period N and

Yk = XkXk e  j o k ... (20)

6.3.9 Running Sum or Accumulation

If x(n)  Xk
y(n)  Yk
n
 1 
Then y(n)=  x(k )
k 
 Yk    jk
 1 e o
 Xk , k  0

... (21)

Proof - consider the running sum

n

y(n) =  x(k )
k 

n 1
y(n) = x ( n)   x(k )
k 

y(n) = x(n)+y(n1)
y(n) y(n) = x(n)
Using the time-shifting and linearity properties, we get
YkYk e  j o k = Xk

 1 
Yk =   jk   Xk
 1 e o 

GATE MASTERS PUBLICATION

 282 Signals & Systems

Study Note

The discrete-time fourier series coefficient Yk of the running sum y(n) =  x(k ) is finite-valued and periodic
k 

only if X0 = 0, that is average value of signal should be zero then only the signal y(n) will be finite.

6.3.10 Conjugation and conjugate symmetery

If x(n)  Xk
Then y(n) = x*(n)  Yk = X*k ... (22)
Taking the complex conjugate of a periodic signal x(n) has the effect of complex conjugation and time
reversal on the corresponding fourier series coefficients.
Proof

1 1
Yk =  y (n)e  jk o n   x* (n)e jk o n
N n  N  N n  N 

*
*
 
1   1 
=   x(n)e jk o n     x(n)e j (  k ) o n 
 N n  N  N n  N 
   
 Xk 

= (Xk)*= X*k

1. When the signal is real

That is x[n] = x*[n]
If x[n]  Xk
and x*[n]  X*k
Using all properties we get
Xk = X*k
Thus for a real time domain signal Xk values are conjugate symmetric. Thus Xk has even magnitude plot and
odd phase plot.

Study Note
Similary if x[n] is imaginary then Xk the DTFS coefficients will be conjugate anti symmetric.

2 . When the signal is real and even

Here x[n] = x*[n] and x[n] = x[n]
 Xk = X*k and Xk = Xk
Using these conditions we get
X*k = Xk = Xk
Thus Xk values are also real and even
3 . when the signal is real and odd
Here x[n] = x*[n] and x[n] = x[n]
 Xk = X*k and Xk =  Xk
Using these conditions we get Xk * = X k

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 283

Thus Xk are imaginary and odd.

4 . If signal is real with DTFS coefficients X k then DTFS coefficients of even and odd part of
signal will be ( X k = Re( X k )+ j Img( X k ))
Here x[n]  Xk
x[n]  x[ n]
And xe[n] =
2

x[n]  x[ n]
And xo[n] =
2

X k  X k
 xe[n] 
2
Since Xk = Xk*

X k  X k*
 xe[n]   Real  X k 
2

X k  X k*
And xo[n]   j Im g  X k 
2

REMEMBER Thus we can club all properties of DTFS coefficients for time doman signal
Time domain DTFS coefficients
signal x[n] Xk
Real  Conjugate symmetric
Imaginary  Conjugate anti symmetric
Even  Even
Odd  Odd
Real and even  Real and even
Real and odd  Imaginary and odd
Imaginary and even  Imaginary and even
Imaginary and odd  Real and odd
The arrow used here is double sided which means for example when time domain signal is
conjugate symmetric then DTFS coefficients will be real and similary for rest all properties
it is both sided

6.3.11 Parseval’s Theorem

If x(n)  Xk
1 | X k |2
Then
N
 | x(n) |2 = 
k  N 
... (23)
n  N 

As in the continuous-time case, the left-hand side of Parseval’s relation is the average power in one period
for the periodic signal x(n). Similarly |Xk|2 is the average power in the kth harmonic component of x(n). Thus,
Parseval’s relation states that the average power in a periodic signal equals the sum of the average powers in all
of its harmonic components. In discrete-time, of course, there are only N distinct harmonic components, and
since the Xk are periodic with period N, the sum on the right-hand side of Eq(23) can be taken over any N

GATE MASTERS PUBLICATION

 284 Signals & Systems

consecutive values of k.
Proof - consider the LHS of Eq. (23),we have

1 2 1 *

N
 | x ( n) |
n N
= N  x ( n ) x ( n)
n N

1  
=  x(n)   X k e jk o n 
N  k N 
n N  

1 1 *  jk o n

=
N
 x(n)  N  X e k 
n N  k N 

 
=  X k*   x ( n )  jk o n 
 n N 
k N  

 X k* X k
= k N

|X k |2
= k N

6.3.12 Duality Theorem

We have seen that using DTFS we transform a discrete periodic x[n] into discrete periodic signal Xk.
x[n]  Xk
1  jo kn
Xk =
N
 x  n e
n  N 
... (24)

And x[n] =  X k e jokn ... (25)

k  N 

Both signals Xk and x[n] are periodic with period N. The format of formulae to find DTFS coefficient is

DTFS coefficient =  Time domain signal e  jo kn

n  N 

If we manipulate eq(25) we get

x[n] 1
=  X k e  jo kn
N N k  N 

Inter changing variable n by k

x[k] 1
 =  X n e  jo kn
N N n  N 

x[ k ]
 Xn 
N

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 285

x[  k ]
Thus are DTFS coefficients of signal Xn , thus the duality theorem is
N
If Xk are DTFS coefficients of x[n], x[n]xk
x[  k ] x[ k ]
Then DTFS coefficients of Xn are Xn 
N N

6.4 Output of discrete system with periodic input

Let us consider a LTI system whose impulse response is h[n] and input is x[n] = Cn where C is constant.
Thus output is
y[n] = x[n]*h[n]


=  h[m]x  n  m 
m 

 
n m
=  h[m]C
m 
= C
n
 h[m]C m

m 

The z-Transform of h[n] is


m
H(z) =  h[n]z
m 

 y[n] = H(z)|z=C . Cn
Thus when input is C then output of LTI system is Cn. H(z)|z=C. From the above analysis we get that
n

LTI system
e jo n h[n]  H(z) e jon .H ( z ) z  e j o

LTI system
X k e jo kn h[n]  H (z)
X k e jo n .H ( z ) jo
z e

x n  X k e jo kn LTI system

 X k e jo n .H ( z ) z  e jo
k N h[n]  H (z) k N

When x[n] is input then output y(n) is

y[n] =   
X k H e jo k e jo kn
k  N 

From the above analysis we get that if input to a LTI system is periodic then output is also periodic with
same period. If DTFS coefficient of x[n] are Xk then output will have DTFS coefficient of Xk H( e j o k ).

6.5 Insights on formulae of DTFS

From eq(6) and (7) we can say that

GATE MASTERS PUBLICATION

286 Signals & Systems

1
 X0 =  x[n], average value of periodic signal x[n]
N k N

 x[o] = 
k  N 
X k , sum of all DTFS coefficients

 If in a question a singal x[n] is given and we have to find phase of Xk then qenerally we can make x[n]
even by shifting it right or left. Thus y[n] = x[n±a] and Yk will have zero phase as y[n] is real and even.
 jao k
Since Yk = Xk e and we can easily find phase of Xk.

 If a signal x[n] is periodic with period N and N is even,

 N
And x[n] =  x  n  
 2
That is signal is half wave symmetric then Xk will be zero for even value of k, because (using time shifting
property)
N
 jo k
2
Xk =  X k e
 Xk = Xke±jk
 Xk + Xke±jk = 0
 Xk = 0 For even values of k
 Similary if N is divisible by 4

 N
And x[n] =  x  n  
 4
Then Xk = 0 For all where k is multiple of 4.
 If x[n] is real and N is even then Xk for k =N/2 will also be real.

1
Since, Xk =  x[n]e jno k
N n  N 

Since x[n] is real and e  jno k will become real when k = N/2.

 If a signal x[n] is not looking like half wave symmetric then it can be shifted on amplitude axis. If after
shifting the signal y[n] = x[n]±a we get it is half wave symmetric then Yk will have property of half wave
symmtric and Yk = Xk for all values except for k = 0 because X0  Y0 as by shifting the signal we have
changed it is average value.
As we have done in CTFS we can write the discrete periodic signal in terms of harmonics of cosine and
sine also. We can also prove that odd signals will have only sine terms and even have only cosine terms.
If any signal is made odd by shifting on amplitude axis then it will show all properties of odd signal except
there will be change in value of DTFS coefficient at k = 0.

Example 6.6
A discrete-time periodic signal x ( n ) is real valued and has a fundamental period N = 5.
Tge nonzero fourier series coefficients for x ( n ) are
X 0 = 1, X 2 = X*2e jπ / 4 , X4 = X*4 = 2e jπ / 3

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 287

Express x ( n ) in the form


x ( n ) = A0 +  Ak sin k n+ k 
k=1

Solution 6.6
The singal x(n) is periodic with period N = 5 and fundamental frequency 0 = 2/N = 2/5.
Using the fourier synthesis , we have
jk o n
x(n) = Xe
k N
k

= X 0  X 2 e j 2o n  X 2 e  j 2o n  X 4e j 4 o n  X 4 e j 4 o n

= X 0  X 2 e j 22  / Nn  X 2 e  j 22  / Nn  X 4 e j 42  / Nn  X 4 e  j 42  / Nn

= 1  e j  / 4 e j 22  /5 n  e  j  / 4 e  j 22  /5 n 2e j  /3e j 42  5/ n  2e  j  /3e 42  /5 n

 e j  4 /5 n  /4   e  j  4  /5n  /4    e j  8 /5n  /3  e  j  8 /5n  /3 

= 1  2   4  
 2   2 

 4   8 5 
= 1  2 cos  n    4 cos  n  
 5 4  5 3 

 4 3   8 5 
= 1  2sin  n    4sin  n  
 5 4   5 6 

= A0   A sin  w n   
k 1
k k k

Example 6.7

Let x ( n ) be a real and odd periodic signal with period N = 7 and fourier series coefficients
X k . Given that
X 15 = j, X 16 = 2 j , X 17 = 3 j
Determine the values of X 0 , X -1 , X -2 , X -3 .
Solution 6.7
The discrete-time fourier series coefficients are periodic with period N, i.e.,
Xk = Xk+N = Xk+2N
Therefor, for N = 7, we have
X 1 = X8 = X15 = j
X 2 = X9 = X16 = 2j
X 3 = X10 = X17 = 3j
Since the given signal x(n) is real and odd, the fourier series coefficients Xk will be purely
imaginary and odd (Xk = Xk). Therefore,
X0 = 0
X 1 = X1
X 2 = X2

GATE MASTERS PUBLICATION

288 Signals & Systems

X3 = X3
Finally, we have X1 = X1 = j
X2 = X2 = 2j
X3 = X3 = 3j

Example 6.8
Suppose we are given the following information about a signal x ( n ) :
1 . x ( n ) is a real and even signal.
2 . x ( n ) has a period N = 10 and fourier coefficients X k .
3 . X 11 = 5.

1 9
4.  n=0
| x(n)|2 = 50.
10
Show that x ( n ) = A cos ( Bn + C ), and specify unmerical values for the constants A, B, and
C.

Solution 6.8
The fourier series coefficients are periodic with period N = 10, we have
Xk = Xk+N = Xk = Xk+10
X 1 = X11 = 5
X1 = X 9
Since, the given signal x(n) is real and even, the fourier coefficients Xk are also real and even. Therefore,
Xk = Xk
X 1 = X1 = X9 = 5
Using Parseval’s relation,
N 1 N 1
1 2
 | x ( n) | = | X k |2
N n0 k 0

1 9 9

 | x(n) |2 =
10 n 0
| X
n 0
k |2  50

8
| X 0 |  | X 1 |  | X 9 |   | X k |2 = 50
2 2

k 2

8
| X 0 |2 25  25   | X k |2 = 50
k 2

8
| X 0 |2 | X k |2 = 0
k 2

Therefor, for one period the fourier coefficients are

5, k  1,9
Xk = 
0, k  0 and 2  k  8
Now using the synthesis equation, we have

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 289
jk o n
x(n) = Xe
k N
k

N 1 9
jk 2  / Nn
=  X ke   X k e jk 2 /10 n
k 0 k 0

= X 1e j 2  /10 n  X 9 e j18  /10 n

Substituting X9 = X1, we get

x(n) = X 1e j 2 /10 n  X 1e j 2 /10 n

= 5e j/5 n  5e  j/5 n

 e j/5 n  e j/5n 
= 10  
 2 

 n 
= 10cos  
 5 
Example 6.9

Each of the two sequences x ( n ) and g ( n ) has a period N = 4, and the corresponding fourier
series coefficients are specified as
x ( n )  y(n)  Yk
1 1
Where X 0 = X3 = X1 = X2 =1
2 2
Y 0 = Y1 = Y2 = Y 3 = 1
Using the multiplication property, determine the fourier series Z k for the signal z ( n ) =
x ( n ) y ( n ).
Solution 6.9
The given signal z(n) = x(n)y(n) is periodic with period N = 4. Using the multiplication property,
we have
z(n) = x(n)y(n)
3

Zk = XY r k r = Zk = X Y
r 0
r k r
r N

Zk = X0Yk+X1Yk1+X2Yk2+X2Yk3
Zk = Yk+2Yk1+2Yk2+Yk3
Since Yk = Yk+N = 1 for all values of k, therefore
Zk = 1 + 2 + 2 + 1 = 6 for all values of k

Example 6.10
When the impulse train


x(n) =   (n- 4k )
r=- 

GATE MASTERS PUBLICATION

290 Signals & Systems

Is the inputs to a particular LTI system with frequency response H ( e jw ), the output of the
system is found to be

5π π
y ( n ) = cos( n+ )
2 4

Determine the values of H ( e jk / 2 ) for k = 0, 1, 2 and 3.

Solution 6.10
The given input signal x(n) is periodic with period N = 4 and frequency 0 = 2/N = 2/4. The
given signal for one period may be written as

1, n  0
x(n) = 
0, n  1, 2,3
= (n)
Therefore the fourier series coefficients are
N 1
1  jk o n
Xk =
N
 x ( n) e
n 0

1 3  jk 2  /4 n 1 3  jk 2 /4 n
=  x ( n ) e =  ( n )e
4 n 0 4 n 0

1
Xk = , for all k
4
we know that the output y(n) of an LTI system to a periodic input is given by,

y(n) = X k H (e jko )e jkon

k N

= X k H (e jk 2 / N )e jk 2 / N
k 0

= X k H (e jk 2 /4 )e jk 2 /4 n
k 0

y(n) = X0H(ej0)ej0n+X1H(ej/2)ej/2n+X2H(ej)ejn+X3H(ej3/2)ej3/2n
1 1 1 1
y(n) = H  e j 0  e j 0 n  H  e j/ 2  e j /2 n  H  e j  e jn  H  e j3/2  e j3/2n
4 4 4 4
Given that

 5      
y(n) = cos  n   = cos  2n  n   = cos  n  
 2 4  2 4 2 4

1 j /2n/4 1  j /2n/4
y(n) = e  e
2 2

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 291

1 j /2 n/4  1 j3 /2 n/4 

y(n) = e  e
2 2

1 j/4 j /2 n  1 /4 j3 /2 n 

y(n) = e e  e e
2 2
Comparing the above equation with Eq.(1), we get
H(ej0) = H(ej) = 0,
H(ej/2) = 2ej/4, H(ej3/2) = 2ej/4
Example 6.11

If the signal is of period N = 8 and it is DTFS coefficients X k are

 kπ   2kπ 
X k = Cos  +sin  
 4   4 
Then find the signal x [ n ].
Solution 6.11
Since Xk is given in sine and cosine, we cannot apply eq(6) to find x[n]. we will do comparison to find x[n].

1  jo kn
Since Xk =
N
 x[n]e
n 8 

2
Since periodic N = 8 , O =
8

e jk /4 e  jk /4 e  j 3k  /4 e j 3k  /4
 Xk =   
2 2 2j 2j
And we will compare above equation with
2
1 4  j .kn
Xk =  x  n  e
8
8 n 3
We get x[1] = 4
x[1] = 4
x[3] = 4j
x[] = 4j
 x[n] = 4j [n+3]4j [n3]+4 [n1]+4 [n+1]
for one period 3 to 4

Example 6.12
Suppose we are given the following information about a periodic signal x ( n ) with period
N = 8 and fourier coefficients X k .
1. X k =  X k 4
2. x (2 n +1) = (1) n
Sketch one period of x ( n )
Solution 6.12
We known that if x(n)  Xk

GATE MASTERS PUBLICATION

292 Signals & Systems

Then, from the frequency shifting property, we have

2 N
j n
(1)nx(n) = ejnx(n) = e N 2 x(n)  XkN/2
Since, it is given that Xk = Xk4, we have
(1)nx(n)  Xk
(1)nx(n)  x(n)
x(n)[1+(1)n]  0
This implies that x(n) = 0, For n = 0,±2, ±4, ±6,
From the given information x(2n+1) = (1)n, we get
x(n) = 1, x(3) = 1, x(5) = 1, x(7) = 1
Therefore, over one period 0 n 7,x(n) is defined as

0, n  0, 2,4,6

x(n) = 1 n  1,5
1 n  3,7

One period of x(n) is shown in figure

xn)



3 7
0 1 2 4 5 6 n



Example 6.13

Let x ( n ) be a periodic signal with period N = 8 and fourier series coefficients X k =  X k 4 .

A signal

 1+  -1n 
y ( n ) =   x  n -1
2 
 
With period N = 8 is generated. Denoting the fourier series coefficients of y ( n ) by Y k , find
a function such that
Y k = f(k)Xk
Solution 6.13
We known that if x(n)  Xk
Then from the frequency shifting property, we have
2 N
j n
(1)nx(n) = ejnx(n) = e N 2 x(n)  XkN/2
In this case N = 8, therefore
(1)nx(n)  Xk4

Since, it is given that Xk = Xk4, we have

(1)nx(n) = Xk

GATE MASTERS PUBLICATION

 Discrete Time Fourier Series 293

(1)nx(n) = x(n)
(1)nx(n) = x(n)
(1)n+1x(n) = x(n)
(1)nx(n1) = x(n1)
Now consider

   1n 
y(n) =   x  n  1
2 
 

1 1 n
= x  n  1   1 x  n  1
2 2
We have already proved that (1)nx(n1) = x(n1), Therefore
y(n) = x(n1)
Using the time-shifting property, we get
Yk = Xkejk2/8 = f(k)Xk
Thus f(k) = ejk2/8

GATE MASTERS PUBLICATION

 294 Signals & Systems

S1 :e
jn /2
 e jn /2u  n  ,

j n / 2
S2 :e  e j 3 n /2
Subjective Practice Problems
j n / 2
S3 :e  e j 5 n /2

1. Each if the two sequences x1[n]and x2[n] has a For each system, determine whether the given
period N = 4, and the corresponding fourier se- information is sufficient to concrete that the sys-
ries coefficients are specified as tem is difinitely not LTI.

x1[n]  ak,
x2[n]  bk, 4. Let

Where
1, 0  n  7
x[n] = 
1 1 0, 8  n  9
a0  a3  a1  a2 =1
2 2
be a periodic signal with fundmental period N=
And b0 = b1 = b2= b3= 1. 10 and Fourier series coefficients ak. Also, let

Using the multilication property, determine the g[n] = x[n]x[n1]

fourier series coefficients ck for the signal g[n] =
(a) Show that g[n] has a fundmental period of 10.
x1[n]x2[n].
(b) Determine the Fourier series coefficients of
2. Determine the output of the filter shown in figure
g[n].
for the following periodic inputs:
5. Consider a causal discrete-time LTI system whose
(a) x1[n] = (1)n
input x[n] and output y[n] related by the following
difference equation:
 3 
(b) x2  n   1  sin  n 
 8 4 1
y n  y  n  1 =x[n]
4
n4k
(c) x3  n  
 1 Find the fourier series representation of the
 k   2  u  n  4k 
output y[n] for each of following inputs:

H (e j  )  3 
(a) x[n] = sin  n
 4 
1

   
(b) x[n] = cos  n   2 cos  n 
2  5  11   5   0      
4  2 

3 12 12 3 3 12 12 3
6. Let x[n] be a periodic sequece with period N and
fourier series representation
3. Consider three discrete-time systems S1,S2 and Ss
whose respective response a complex input ejn/2 jk  2  / N  n .

are specified as
x[n]= 
k  N 
ak e

GATE MASTERS PUBLICATION

 Discrete Time Fourier Seriess 295

The fourier series coefficients for each of the  NIM  1

following signals can be expressed in terms of ak.  N
Derive the expressions.

r 0
x  n  r  = 0 for all n,
 M
(a) x[n-n0]
Then ak = 0 for every value of k that is a mul-
(b) x[n]x[n1] tiple of M.

 N
(c) x[ n]  x  n  (assume that N is even) 8. Suppose we are given the following information
 2 
about a periodic signal x[n] period 8 and fourier
coefficients ak:
 3 
(d) sin  n  (assume that N is even; note that 1. ak =  ak 4.
 4 
2. x[2n + 1] = (1)n.
this signal is periodic with period N/2 )
Sketch one period of x[n].
(e) x*[n]
9. Let x[n] be a real periodic signal with period N
(f) (1)n x[n] (Assume that N is even)
and Fourier coefficients ak.
(g) (1)n x[n] (Assume that is odd; note that
(a) Show that if N is even, at least two of the fourier
this signal is periodic with period 2N)
coefficients within one period of ak are real.

 x  n  , n even (b) Show that if N is odd, at least one of the fourier

(h) y[n]   coefficients within one period of ak is real.
0, n odd
10. Evaluate the time-domain signals represented by
7. Let x[n] be a periodic sequence with period N and the following DTFS coefficients:
fourier series representation
(a) X k  cos 
 8 
jk  2  / N  n .
k
x[n]=  ak e  21 
k  N 

(b) X k  cos 
10   4 
k   j 2 sin  k
(a) Suppose that N is even and  19   19 


 N (c) X k  m
x[n]=  x  n 
 2 
for all n,   1
m
  k  2m  2  k  3m  

Show that ak = 0 for all even integers k. (d) Xk as depicted in Figure.

(b) Suppose that N is divisible by 4. Show that if

|Xk|
 N
x[n]=  x  n 
 4 
Then ak = 0 for every value of k that is a mul- 1
tiple of 4.
10 11 4 3 0 34 1011 k
(c)More generally, suppose that N is divisible by
an integer M. Show that if

GATE MASTERS PUBLICATION

 296 Signals & Systems

 6 
Xk (a) x(n) = cos  n 
 17 3


2  14   10 
 4 11 (b) x(n) = 2sin  n   cos  n
4 0 3 10 k
 19   19 
 
2
 m
(c) x(n)  
m
(1)  n  2m  n  3m 
11. One period of the DTFS coefficients of a signal is
13. Consider a dicrete-time LTI system whose fre-
given by
quency response is
k
1  1, |  |  / 8
Xk=   , 0  k  9 H(ej)= 
2
0,  / 8 |  | 
Find the time-domain signal x(n) assuming N = 10
Show that if the input x(n) to this system has a
period N = 3, the output y(n) has only nonzero
12.Evaluate the DTFS coefficients of the following fourier series coefficient per period.
signals:

GATE MASTERS PUBLICATION