Digital Signal Processing Reference
In-Depth Information
(
we will use these two equations for the DFT-IDFT pair in the remaining pages
):
N
−
1
N
−
1
1
N
1
N
X(k)e
j(
2
π/N)kn
X(k)W
−
kn
,
x(n)
=
=
0
≤
n
≤
N
−
1
(3.79)
k
=
0
k
=
0
N
−
1
N
−
1
x(n)e
−
j(
2
π/N)kn
x(n)W
kn
,
X(k)
=
=
0
≤
k
≤
N
−
1
(3.80)
n
=
0
n
=
0
In Figure 3.29b, we have shown the DFT as a subset of the Fourier series
coefficients
X
p
(k)
,for
k
−
1
)
. But we can choose any other
N
consecutive samples as the DFT of
x(n)
[e.g.,
=
0
,
1
,
2
,...,(N
−
[
(N
−
1
)/
2]
≤
n
≤
[
(N
−
1
)/
2]], so we will use the notation
N
=
n
modulo
N
to denote that
n
ranges
over one period of
N
samples.
Given a nonperiodic discrete-time function
x(n)
, we constructed a mathe-
matical artifact
x
p
(n)
and derived the Fourier series representation for it and
also derived its inverse to get
x
p
(n)
. Then we defined the DFT and IDFT as
argued above so that we could determine the frequency response of the non-
periodic function as samples of the DTFT
X(e
jω
)
at
N
equally spaced points
ω
k
=
(
2
π/N)k
. We know
x(n)
is nonperiodic, but since
X(e
jω
)
is periodic with
aperiod2
π
,
X(e
jω
k
)
=
X
p
(k)
is periodic with a period
N
, so one can choose
the range
n
in Equations (3.79) and (3.80).
The two equations for the DFT and IDFT give us a numerical algorithm
to obtain the frequency response at least at the
N
discrete frequencies, and
by choosing a large value for
N
, we get a fairly good idea of the frequency
response for
x(n)
.
6
Indeed, we show below that from the samples of
X(k)
,we
can reconstruct the DTFT of
x(n)
=
N
X(e
jω
)
, which is a function of the contin-
uous variable
ω
. This is the counterpart of Shannon's reconstruction formula to
obtain
x(t)
from its samples
x(n)
, provided
x(n)
is bandlimited and the sam-
pling period
T
s
<(π/ω
b
)
. There are similar conditions to be satisfied in deriving
the formula in the frequency domain, to reconstruct
X(e
jω
)
from its samples
X(e
jω
k
)
=
=
X(k)
.
3.6.4 Reconstruction of DTFT from DFT
First we consider the DTFT of
x(n)
and substitute
x(n)
by the formula (3.79)
for finding the IDFT of
X(k)
as explained below:
1
N
X(k)e
j(
2
πkn/N)
e
−
jωn
N
−
1
N
−
1
N
−
1
X(e
jω
)
x(n)e
−
jωn
=
=
n
=
0
n
=
0
k
=
0
N
−
1
N
−
1
1
N
e
j(
2
πkn/N)
e
−
jωn
=
X(k)
(3.81)
k
=
0
n
=
0
6
Later we will discuss what is known as the “picket fence effect,” because of which we may not get
a fairly good idea of the frequency response.
Search WWH ::
Custom Search