0% found this document useful (0 votes)
131 views

ECEN 314: Signals and Systems

This document provides lecture notes on discrete-time Fourier series. It defines periodic discrete-time signals and their fundamental periods. It then introduces the discrete-time Fourier series representation of periodic signals as a linear combination of complex exponentials. Finally, it answers two questions: 1) which periodic signals have a Fourier series representation, and 2) how to calculate the Fourier coefficients.

Uploaded by

Abdul Ahad
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
131 views

ECEN 314: Signals and Systems

This document provides lecture notes on discrete-time Fourier series. It defines periodic discrete-time signals and their fundamental periods. It then introduces the discrete-time Fourier series representation of periodic signals as a linear combination of complex exponentials. Finally, it answers two questions: 1) which periodic signals have a Fourier series representation, and 2) how to calculate the Fourier coefficients.

Uploaded by

Abdul Ahad
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

ECEN 314: Signals and Systems

Lecture Notes 10: Discrete-Time Fourier Series

Reading:

• Current: SSOW 3.6-3.7

• Next: SSOW 3.9

1 Periodic DT signals
A DT signal x[n] is said to be periodic if there exists a positive integer N such that

x[n + N ] = x[n], ∀n ∈ Z

where N is called a period. The smallest such N is called the fundamental period and is
usually denoted as N0 . The fundamental frequency is defined as

Ω0 :=
N0
If N0 is the fundamental period of x[n], then any positive integer multiple N = kN0 is a
period of x[n].
The real (complex) sinusoid x[n] = cos(Ωn + θ) (x[n] = ej(Ωn+θ) ) is periodic if and only
if |Ω| is a rational multiple of π of the form
k
|Ω| = 2π.
N0
Furthermore, if k and N0 are co-prime, the fundamental period of the signal is given by N0 .
Examples:

• x[n] = cos(n).
π

• x[n] = cos 2
n
.

• x[n] = cos 3π

2
n .

• x[n] = cos(2πn).

Combining periodic DT signals:

1
• Combining periodic DT signals always results in a periodic DT signal. A period of the
combined signal is given by the lowest common multiple (LCM) of the periods of the
individual signals.

• It is possible that the fundamental period of the overall signal is smaller than the LCM
of the fundamental periods of the individual signals.

2 Discrete Time Fourier Series


Let x[n] be a periodic DT signal with fundamental period N0 . Similar to CT signals, we
want to write x[n] as a linear combination of complex exponentials, e.g.

X
x[n] = ak ejkΩ0 n ,
k=−∞

where φk [n] = ejkΩ0 n are periodic complex sinusoids for all integer k.
The key difference is that, the CT signals φk (t) = ejkω0 t are distinct for all values k. For
DT signals, this is not the case. Since the time variable n must be an integer, we have

φk+N0 [n] = ej(k+N0 )Ω0 n = ejkΩ0 n · ejN0 Ω0 n = ejkΩ0 n · ej2πn = ejkΩ0 n = φk [n]

Therefore, there are only N0 distinct signals in the set {φk [n] | k ∈ Z}.
Implications: To uniquely determine the coefficients ak , we need to specify N0 different
φk [n]. In theory, we can pick any N0 distinct φk [n]. In practice, it is customary to choose
φk [n] for k = 0, 1, . . . , N0 − 1. The sum in the Fourier series representation is thus written
as
N
X0 −1

x[n] = ak ejkΩ0 n .
k=0

Two questions:

1) What periodic DT signals will have such Fourier series representation?

2) How do we find {ak }?

2.1 Answering Question 1


For any DT periodic signal with period N , there is a Fourier series representation. This can
be seen by choosing Ω0 = 2π/N writing the Fourier series representation for n = 0, 1, . . . , N −

2
1 as
N
X −1
x[0] = ak
k=0
N
X −1
x[1] = ak ejkΩ0
k=0
N
X −1
x[2] = ak ej2kΩ0
k=0
..
.
N
X −1
x[N − 1] = ak ej(N −1)kΩ0 .
k=0

Note that this is a system of N linear equations with N unknowns a0 , a1 , . . . , aN −1 . In the


matrix-vector form, we have
    
x[0] 1 1 ··· 1 a0
 x[1]  
   1 ejΩ0 ··· ej(N −1)Ω0 
 a1 

 x[2]  
 = 1 ej2Ω0 ··· ej2(N −1)Ω0 
 a2 

 ..   ..  .. 
 .   .  . 
x[N − 1] 1 ej(N −1)Ω0 · · · ej(N −1)(N −1)Ω0 aN −1
The matrix is a scalar multiple of a unitary matrix (i.e., a complex rotation) and is known
to be invertible for any positive integer N . Hence, given x[0], x[1], . . . , x[N − 1], we can
uniquely determine the Fourier series coefficients a0 , a1 , . . . , aN −1 by “inverting” the linear
system equations.
Note that the answer for the DT case is very different from the CT case where only signals
that satisfy the Dirichlet conditions will have Fourier series representations. The reason is
that for the CT case, the summation in the synthesis equation is an infinite series, while for
the DT case it is only over a finite number of terms.

2.2 Answering Question 2


Recall the finite geometric sum given by
b 
X
n b − a + 1,  α=1
α = 1 a b+1
1−α
α − α 6 1.
, α=
n=a

Let α = ejkΩ0 = ejk2π/N . Then, α = 1 when k is an integer multiple of N and α 6= 1


otherwise. Thus, we have
N −1 
X
jkΩ0 n N, if k is an integer multiple of N
e = 1−ejkΩ0 N 1−ejk2π
1−ejkΩ0
= 1−ejkΩ0 = 0, otherwise.
n=0

3
Now, we assume that
N
X −1
x[n] = ak ejkΩ0 n .
k=0
−jmΩ0 n
Then, we multiply both sides by e , for some fixed m ∈ {0, . . . , N − 1}, and sum over
n = 0, . . . , N − 1. This gives
−1 −1 −1
N N N
!
X X X
x[n]e−jmΩ0 n = ak ejkΩ0 n e−jmΩ0 n
n=0 n=0 k=0
−1 −1
N N
!
X X
= ak ej(k−m)Ω0 n
k=0 n=0

Note that
N −1 
X
j(k−m)Ω0 n N, if k = m
e =
0, if k = 0, . . . , m − 1, m + 1, . . . , N − 1
n=0

We thus have
N
X −1
x[n]e−jmΩ0 n = N am
n=0

giving
N −1
1 X
am = x[n]e−jmΩ0 n
N n=0
1 X
= x[n]e−jmΩ0 n
N
n∈hN i

for m = 0, . . . , N − 1, where hN i denotes any set of N consecutive integers.


Periodic extension of ak : As mentioned previously, for DT signals φk [n] = ejkΩ0 n ,
φk+N [n] = φk [n]. Therefore, analysis equation above implies that ak+N = ak and one can
think of ak as a periodic sequence with period N . In comparison, CTFS coefficients are
generally aperiodic.
With this periodic extension of ak , we have the following general DTFS pair (for any
periodic x[n] with period N ):

X
x[n] = ak ejkΩ0 n (Synthesis equation)
k∈hN i
1 X
ak = x[n]e−jkΩ0 n (Analysis equation)
N
n∈hN i

4
3 Examples
Example 1: Assume that N ≥ 2N1 + 1 and let x[n] be the periodic signal (with period N )
given by 
1 if 0 ≤ n ≤ N1

x[n] = 1 if N − N1 ≤ n < N

0 otherwise.

Find the FS coefficients of the periodic DT square wave x[n].


Answer: By the analysis equation,
1 X
ak = x[n]e−jkΩ0 n
N
n∈hN i
N1
1 X n
= e−jkΩ0 .
N n=−N
1

When k is an integer multiple of N , we have e−jkΩ0 = e−sk(2π/N ) = 1 and hence


N1
1 X 2N1 + 1
ak = 1= = d,
N n=−N N
1

where d is the duty cycle of the periodic DT square wave. When k is not an integer multiple
of N , then we have e−jkΩ0 6= 1 and we can use the geometric sum formula to get
N1
1 X n
ak = e−jkΩ0
N n=−N
1

1e jkΩ0 N1
− e−jkΩ0 (N1 +1)
=
N 1 − e−jkΩ0
1e jkΩ0 N1
− e−jkΩ0 (N1 +1) ejkΩ0 /2 e−jkΩ0 /2
=
N 1 − e−jkΩ0 ejkΩ0 /2 e−jkΩ0 /2
jkΩ0 (N1 −1/2) −jkΩ0 (N1 +1/2) −jkΩ0 /2
1e −e e
=
N ejkΩ0 /2 − e−jkΩ0 /2 e−jkΩ0 /2
sin(πkd)
= .
N sin(πk/N )

The 3rd and 4th steps make use of a valuable trick related to the following general observation

ea − e−b = ea − e−b e(b−a)/2 e−(b−a)/2




= e(a+b)/2 − e−(a+b)/2 e−(b−a)/2




= 2j sin((a + b)/2))e−(b−a)/2 .

5
4 Properties of DTFS
The first set of properties can be proven in a manner nearly identical to the analagous
property for CTFS.
Property 1 (Linearity). Let x[n] and y[n] be two periodic DT signals with the same period
DT F S DT F S
N . Suppose that x[n] ←−−→ ak and y[n] ←−−→ bk . Then
DT F S
αx[n] + βy[n] ←−−→ αak + βbk
DT F S
Property 2 (Time Reversal). Suppose that x[n] ←−−→ ak . Then
DT F S
x[−n] ←−−→ a−k
DT F S
Property 3 (Even and Odd Symmetry). Suppose that x[n] ←−−→ ak . If x[n] is even, then
ak is also even, i.e., a−k = ak . If x[n] is odd, then ak is also odd, i.e., a−k = −ak .
DT F S
Property 4 (Conjugation). Suppose that x[n] ←−−→ ak . Then
DT F S
x∗ [n] ←−−→ a∗−k
DT F S
Property 5 (Conjugate Symmetry). Suppose that x[n] ←−−→ ak and x[n] is real. Then
a−k = a∗k
DT F S
Remark 1. The DTFS pair, x[n] ←−−→ ak , satisfies the following symmetry conditions:
1. If x[n] is real, then Re{ak } is even, Im{ak } is odd, |ak | is even, and ∠ak is odd.
2. If x[n] is real and even, then ak is real and even
3. If x[n] is real and odd, then ak is purely imaginary and odd.
DT F S
Property 6 (Time Shift). Suppose that x[n] ←−−→ ak . Then
DT F S
x[n − n0 ] ←−−→ ak e−jkΩ0 n0
The next set of properties are similar to properties of the CTFS but some have important
differences.
Property 7 (Multiplication). Let x[n] and y[n] be two periodic DT signals with the same
DT F S DT F S
period N . Suppose that x[n] ←−−→ ak and y[n] ←−−→ bk . Then
DT F S
x[n]y[n] ←−−→ N (ak ~ bk )
where
1 X
ak ~ bk := al bk−l .
N
l∈hN i

6
Property 8 (Parseval’s Relation). Let x[n] be a periodic DT signal with period N . Suppose
DT F S
that x[n] ←−−→ ak . Then
1 X X
|x[n]|2 = |ak |2 .
N
n∈hN i k∈hN i

Property 9 (Periodic Convolution). Let x[n] and y[n] be two periodic DT signals with the
DT F S DT F S
same period N , and let z[n] = x[n] ~ y[n]. Suppose that x[n] ←−−→ ak and y[n] ←−−→ bk .
Then
DT F S
z[n] ←−−→ ak bk .

Property 10 (Duality). A very important property of the DTFS is that


DT F S DT F S
x[n] ←−−→ ak = y[k] ←−−→ bm = x[−m].

In words, this means that computing the DTFS of the DTFS of a signal returns the original
signal with a time reversal.
DT F S
Proof. Suppose x[n] is periodic with period N and assume that x[n] ←−−→ ak and y[k] = ak
for all integer k. Since ak is periodic with period N , we can also compute the DTFS pair
DT F S
y[k] ←−−→ bm . In this case, we find that
1 X
bm = y[k]e−jmΩ0 k
N
k∈hN i
 
1 X
1
X
= x[n]e−jkΩ0 n  e−jmΩ0 k
N N
k∈hN i n∈hN i
1 X 1 X −jkΩ0 (m+n)
= x[n] e
N N
n∈hN i k∈hN i
1 X
= x[n]δ[m + n]
N
n∈hN i

= x[−m].

You might also like