Professional Documents
Culture Documents
DFT PDF
DFT PDF
j2
N kn
, k = 0, 1, 2, ...
(6.1)
All these signals have frequencies is that are multiples of the some fundamental
frequency, 2
N , and thus harmonically related.
These are two important distinction between continuous time and discrete time
complex exponential. The rst one is that harmonically related continuous time
complex exponential ej0 kt are all distinct for dierent values of k, while there
are only N dierent signals in the set.
The reason for this is that discrete time complex exponentials which dier in
frequency by integer multiple of 2 are identical. Thus
2
where
=
X[k]
N
1
x
[n]ej N kn
(6.3)
n0
In equation (6.2) and (6.3) we can sum over any consecutive N values. The
equation (6.2) is synthesis equation and equation (6.3) is analysis equation.
Some people use the faction 1/N in analysis equation. From (6.3) we can see
easily that
x
[k] = x
[k + N ]
Thus discrete Fourier series coecients are also periodic with the same period N .
Example 1:
{
x[n]} = {cos
x
[n] =
4
},
5
2
1 j 2 .2
(e 5 + ej 5 .2 )
2
= 5 and X[2]
So, X[2]
= 52 , since the signal is periodic with period 5, coe2
cients are also periodic with period 5, and x
[k], 1 k 1.
FIGURE
Now we show that substituting equation (6.3) into (6.2) we indeed get x
[n].
N
1
j 2
N kn =
X[k]e
k=0
N
1
k=0
N 1
2
2
1
x
[m]ej N km ej N kn
N m=0
N
1
x
[m]
m=0
N 1
1 j 2 (nm)k
e N
N
(6.4)
k=0
{X[n]}
X[k]
where LHS represents the signal and RHS its DFS coecients
1. Periodicity DFS coecients:
As we have noted earlier that DFS Coecients {X[k]}
are periodic with period
N.
2. Linearity of DFS:
If
{
x[n]} {X[k]}
{
y [n]} {Y [k]}
j 2
N ln
{e
l]}
x
[n]} {X[k
(6.5)
(6.6)
To prove the rst equation we use equation (6.3). The DFS coecients are
given by
N
1
2
x
[n m]ej N kn
n=0
let n m = l, we get
=
N 1m
x
[l]ej N k(m+l)
lm
since x
[l] is periodic we can use any N consecutive values, then
2
= ej N km
N
1
x
[l]ej N kl
l=0
= ej
2
N km
X[n]
We can prove the relation (6.6) in a similar manner starting from equation (6.3)
4. Duality:
From equation (6.2) and (6.3) we can see that synthesis and analysis equation
dier only in sign of the exponential and factor 1/N . If {
x[n]} is periodic with
period N , then {X[k]}
is also periodic with period N . So we can nd the
discrete fourier series coecients of X[n]
sequence.
From equation (6.2) we see that
Nx
[n] =
N
1
j 2
N kn
X[k]e
k=0
Thus
Nx
[n] =
N
1
j 2
N kn
X[k]e
k=0
N
1
n=0
j 2
N kn
X[n]e
comparing this with (6.3) we see that DFS coecients of {X[n]}
are {N x
[k]},
the original periodic sequence is reversed in time and multiplied by N . This is
known as duality property. If
then
{
x[n]} {X[k]}
(6.7)
{X[n]}
{N x
[k]}
(6.8)
j 2
N kn
x
[n]e
N 1
n=0
j 2
N (k)n
x
[n]e
n=0
[k]
= X
6. Time reversal:
{
x[n]} {X[k]}
From equation (6.3) we have the DFS coecient
N
1
x
[n]ej N kn
n0
putting m = n we get
0
x
[m]ej N km
m=(N 1)
Since x
[m] is periodic, we can use any N consecutive values
=
N
1
x
[m]ej N km
m=0
= X[k]
7. Symmetry properties of DFS coecient:
In the last chapter we discussed some symmetry properties of the discrete time
Fourier transform of aperiodic sequence. The same symmetry properties also
hold for DFS coecients and their derivation is also similar in style using linearity, conjugation and time reversal properties DFS coecients.
8. Time scaling:
Let us dene
x[n/m],
x
(m) [n] =
if n is multiple of m
0, if n is not a multiple of m
sequence {
x(m) [n]} is obtained by inserting (m 1) zeros between two consecutive values of x
[n]. Thus {
x(m) [n]} is also periodic, but period is mN . The
DFS coecients are given by
mN
1
x
(m) [n]ej mN kn
n=0
putting n = lm + r, 0 l N 1, 0 r < m
=
N
1
2 k
m (lm)
x
[l]ej N
l=0
n
x
[k]
k=
{
y [n]} will be bounded and periodic only if the sum of terms of x
[n] over one
N
1
= 0. Assuming this
period is zero, i.e.
x
[n] = 0, which is equivalent to X[0]
to be true
n=0
n
x
[k]
1 ej
k=
1
2 k
N
X[h]
N 1
2
1
X3 [k]ej N kn
N
k=0
1
N
N
1
1 [k]X
2 [k]ej 2
N kn
X
k=0
1 [k] in terms of x
substituting for X
n we get
=
N 1 N 1
2
1
2 [k]ej 2
N kn
x
1 [m]ej N km X
N
m=0
k=0
N
1
m=0
N
1
x
1 [m]
N 1
2
1
X2 [k]ej N (nm)k
N
k=0
x
1 [m]
x2 [n m]
(6.15)
m=0
1 [k]X
2 [k]}
x
1 [m]
x2 [n m] {X
m=0
The sum in the equation (6.15) looks like convolution sum, except that the
summation is over one period. This is known as periodic convolution. The
resulting sequence {
x3 [n]} is also periodic with period N . This can be seen
from equation (6.15) by putting m + N instead of m.
The Duality theorem gives analogous result when we multiply two periodic
sequences.
N 1
1
X1 [l]X2 [k l]
x2 [n]}
{
x1 [n]
N
l=0
1 [k]} and
The DFS coecients are obtained by doing periodic convolution of {X
{X2 [k]} and multiplying the result by 1/N. We can also prove this result directly
by starting from the analysis equation.
The periodic convolution has properties similar to the aperiodic (linear convolution). It is cumulative, associative and distributes over additions of two signals.
The properties of DFS representation of periodic sequence are summarized in
the table 6.2
e
1
X[k]
2
{x[n]
n x[n 1]}
x
[m]
10.
9.
1ej
m=
(periodic only if x
[0] = 0)
11.
12.
13.
14.
15.
16.
17.
N 1
m=0
x
[m]
y [n m]
Y [k]}
{X[k]
N
1
y [k l]}
X[l]
{
x[n]
y [n]}
{ N1
{Re[
x[n]]}
x[n]]}
{jIm [
{
xe [n]} = { 12 (
x[n] + x
[n])}
1
{
xo [n]} = { 2 (
x[n] x
[n])}
If {
x[n]} is real then
e [k]} = { 1 (X[k]
+X
[k])}
{X
2
1
[k])}
{X0 [k]} = { 2 (X[k] X
{Re(X[n])}
{jIm [
x[k]]}
[k]
X[k] = X
Re[X[k]] = Re[X[k]]
Im[X[k]]
= Im[X[k]]
|X[k]| = |X[k]|
= X[k]
X[k]
l=0
N 1
2
2
X[k]
k + 2 l
N
N
l= k=0
2
2 l
X[l]
N
N
l=
as X[h]
are periodic with period N .
[n rN ]
r=
then X[k]
N
1
x
[n]ej N kn
n=0
2 k
2
N
N
k=
x[n mN ]
(6.16)
m=
Note that {
x[n]} dened by equation (6.16) will always be a periodic sequence
with period N, whether {x[n]}is of nite length N or not. But when {x[n]} has
nite length N, we can recover the sequence {x[n]} from {
x[n]} by dening
x
[n], 0 n N 1
(6.17)
x[n] =
0,
otherwise
This is because of {x[n]} has nite length N , then there is no overlap between
terms x[n] and x[n mN ] for dierent values of m = 0.
Recall that if
n = kN + r, where 0 r N 1
then
n modulo N = r,
i.e. we add or subtract multiple of N from n until we get a number lying between
0 to N 1. We will use ((n))N to denote n modulo N . Then for nite length
sequences of length N equation (6.16) can be written as
x
[n] = x[((n))N ]
(6.18)
with it.
This periodic sequence has discrete Fourier series coecients {X[k]},
which are
also periodic with period N . From equations (6.2) and (6.3) we see that we
need values of x
[n] for 0 n N 1 and X[k]
for 0 k N 1. Thus we
dene discrete Fourier transform of nite length sequence {x[n]} as
X[k],
0k N 1
X[k] =
0,
otherwise
where {X[h]}
is DFS coecient of associated periodic sequence {
x[n]}. From
{X[k]}we can get {X[h]}
by the relation.
= X[((k))N ] = X[(k modulo N )]
X[k]
then from this we can get {
x[n]} using synthesis equation (6.2) and nally {x[n]}
using equation (6.17). In equations (6.2) and (6.3) summation interval is 0 to
N 1, we can write X[k] directly in terms of x[n], and x[n] directly in terms of
X[k] as
X[k] =
N
1
x
[n]ej N kn =
n=0
N
1
x[n]ej N kn ,
0k N 1
n=0
0,
N
1
N 1
2
1
j 2
1
N hn =
X[k]e
X[k]ej N hn ,
N
x[n] = N k=0
k=0
0,
otherwise
0N N 1
otherwise
WN = ej N
(6.19)
with this notation, DFT analysis and synthesis equations are written a follows
Analysis equation:
X[k] =
N
1
x[n]WNkn , 0 k N 1
(6.20)
n=0
Synthesis equation:
x[n] =
N 1
1
X[k]WNkn , 0 n N 1
N
(6.21)
k=0
N
1
x[n]ej N kn =
x[n]ej N kn
n=
k=0
(6.22)
2
N k,
k = 0, 1..N 1.
and
y[n] =
0nN 1
0,
otherwise
(6.23)
The shift dened in equation (6.23) is known as circular shift. This is similar
to a shift of sequence in a circular register.
FIGURE
and
(6.24)
(6.25)
4. Duality
We have the duality for the DFS coecient given by {X[n]}
{N X[k]},
retaining one period of the sequences the duality property for the DFT coecient
will become
{X[n]} {N x[((k))N , 0 k N 1}
5. Symmetry properties
We can infer all the symmetry properties of the DFT from the symmetry properties of the associated periodic sequence {
x[n]} and retaining the rst period.
Thus we have
{
x [n]} {X [((k))N ], 0 k N 1}
and
{X [((n))N ], 0 n N 1} {X [k]}
11
1
(x[n] + x [((n))N ]), 0 n N 1
2
1
(x[n] x [((n))N ]), 0 n N 1
2
0
n=0
((n))N =
N n, 1 n N 1
xop [n] = x
0 [n] =
Since
Re(x[0]),
n=0
xep [n] = 1
(x[n] x [N n]), 1 n N 1
2
jIm(x[0]),
n=0
xop [n] = 1
(x[n] x [N n]),
1nN 1
2
x[n] = xep [n] + xop [n]
(6.26)
(6.27)
{xep [n]} and {xop [n]} are referred to as periodic conjugate symmetric and periodic conjugate anti-symmetric parts of x[n]. In terms if these sequence the
symmetric properties are
{Re(x[n])}
{Xep [k]}
{jIm(x[n])}
{xep [n]}
{xop [n]}
{Xop [k]}
{Re(X[k])}
{jIm(X[k])}
6. Circular convolution
We saw that multiplication of DFS coecients corresponds of periodic convolution of the sequence. Since DFT coecients are DFS coecients in the interval,
0 k N 1, they will correspond to DFT of the sequence retained by
periodically convolving associated periodic sequences and retaining their rst
period.
x
[n] = x[((n))N ]
y[n] = y[((n))N ]
Periodic convolution is given by
z[n] =
N
1
x
[k]
y [n k]
k=0
N
1
k=0
12
and then
z[n] =
Since ((k))N = k,
z[n] =
z[n],
0nN 1
0,
otherwise
0k N 1
N
1
we get
x[k]y[((n k))N ],
0nN 1
(6.28)
l=0
The convolution dened by equation (6.28) is known as N-point-circular convolution of sequence {x[n]} and {y[n]}, where both the sequence are considered
sequence of length N. From the periodic convolution property of DFS it is clear
that DFT of {z[n]} is {X[k]Y [k]}. If we use the notation
{x[n]}
N {y[n]}
to denote the N point circular convolution we see that
{x[n]}
N {y[n]} {X[h]Y [k]}
(6.29)
1
{X[k]}
N {Y [k]}
N
(6.30)
Properties of the Discrete Fourier transform are summarized in the table 6.2
13
Table 6.2
Finite length sequence (length N)
1. {x[n]}
2. a{x[n]} + b{y[n]}
3. {X[n]}
4. {x[((n m))N ]}
5. {WNln x[n]}
6. {x[n]}
N {y[n]}
7. {x[n]y[n]}
8. {x [n]}
9. {x [((n))N ]}
10. {Re x[n]}
11. {jIm(x[n])}
12. {Xep [n]}
13. {Xop [n]}
14 If {x[n]} is real sequence
x[k]y[n k]
k=
L1
x[k]y[n k]
(6.31)
k=0
n=
14
w[n]ejwn
W (ejw ) =
w[mN + l]ejw(mN +l)
l=0 m=
If we take w =
2
N k,
we see that
j 2
N k
W (e
)=
N
1
W [l + mN ]ej N l
l=0 m=
2
Comparing this with the DFT equation (6.), we see that W (ej N k ) can be seen
as DFT coecients of a sequence
v[l] =
w[l + mN ],
0l N 1
(6.32)
m=
0l N 1
However, if the length of {W [n]} is greater than N , v[l] may not be equal to
w[l] for all values of l.
The sequence {z[n]} in equation (6.31) has the discrete Fourier transform
Z(ejw ) = X(ejw )Y (ejw )
The N-point DFT of {z[n]} sequence is
Z[k]
= Z(ej N k )
2
2
= X(ej k)Y (ej k)
N
N
= X[k]Y [k]
where {X[k]} and {Y [k]} are N-point DFTs of {x[n]} and {y[n]} respectively.
The sequence resulting as the inverse DFT of Z[k] is then by equation (6.32).
z[n + mN ],
0nN 1
v[n] = m=
0,
otherwise
From the circular convolution property of the DFT we have
{v[n]} = {x[n]}
N {y[n]}
Thus, the circular convolution of two-nite length sequences can be viewed as
linear convolution, followed time aliasing, dened by equation (6.32). If N is
greater than or equal to (L + M l), then there will be no time aliasing as
the linear convolution produces a sequence of length (L + M l). Thus we can
use circular convolution for linear convolution by padding sucient number of
zeros at the end of a nite length sequence. We can use DFT algorithm for
calculating the circular convolution.
15