Scribd
Upload
145 views3 pages

DFT & Idft

This document discusses frequency domain sampling and the discrete Fourier transform (DFT). It explains that sampling the frequency spectrum of a finite-duration sequence x(n) of length L uniquely represents x(n), while sampling an infinite sequence results in aliasing. The DFT takes a sequence x(n) of length L and maps it to N frequency samples, where N >= L. The inverse DFT transforms the frequency samples back to the original sequence. Zero padding the input does not provide extra spectral information but improves the Fourier transform display.

Uploaded by

Istiaque Ahmed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
145 views3 pages

DFT & Idft

This document discusses frequency domain sampling and the discrete Fourier transform (DFT). It explains that sampling the frequency spectrum of a finite-duration sequence x(n) of length L uniquely represents x(n), while sampling an infinite sequence results in aliasing. The DFT takes a sequence x(n) of length L and maps it to N frequency samples, where N >= L. The inverse DFT transforms the frequency samples back to the original sequence. Zero padding the input does not provide extra spectral information but improves the Fourier transform display.

Uploaded by

Istiaque Ahmed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Sec. 5.

1 Frequency Domain Sampling: The Discrete Fourier Transform 399

The periodic sequence x , ( n ) , corresponding to the frequency samples X (2nk/h'),


k = 0 , 1 , .. . , N - 1, can be obtained from either (5.1.4) or (5.1.8). Hence

where the factor 1/(1 - a N ) represents the effect of aliasing. Since 0 < a < 1, the
aliasing error tends toward zero as N -+ oo.
For o = 0.8, the sequence x ( n ) and its spectrum X ( w ) are shown in Fig. 5.4a
and b, respectively. The aiiased sequences x,(n) for N = 5 and N = 50 and the
corresponding spectral samples are shown in Fig. 5 . 4 ~ and d, respectively. We note
that the aliasing effects are negligible for N = 50.
If we define the aliased finite-duration sequence x(n) as
OsnzN-1
otherwise
then its Fourier transform is
N- 1 H-1

Note that although i ( w ) # X ( L L ) ) ,the sample values at cuk = 2nk/N are identical.
That is,

5.1.2 The Discrete Fourier Transform (OFT)

The development in the preceding section is concerned with the frequency-domain


sampling of an aperiodic finite-energy sequence x ( n ) . In general, the equally
spaced frequency samples X ( 2 r k / N ) , k = 0 , l . . . . , N - 1,do not uniquely represent
the original sequence x ( n ) when x ( n ) has infinite duration. Instead, the frequency
samples X ( 2 n k / N ) , k = 0, 1. . . . , N - 1, correspond to a periodic sequence x , ( n )
of period N , where x,(n) is an aliased version of x ( n ) , as indicated by the relation
in (5.1.4), that is,

When the sequence x ( n ) has a finite duration of length L 5 N, then x , ( n )


is simply a periodic repetition of x ( n ) , where x,(n) over a single period is
400 The Discrete Fourier Transform: Its Properties and Applrcatiaos Chap. 5

Figure 5.4 f a ) Plot of sequence A - i n ) = ( O . ) i i " r t ( r i ) : ( h ) its Fourier transform (magnitude


only): ( c ) effect oI aliasing w ~ t hN = 5: ( d ) reduced effect of aliasing with h' = 5 0 .

given as

Consequently. the frequency samples X ( 2 x k / N ) , k = 0. 1.. . . . N - 1 . uniquely


represent the finite-duration sequence x ( , r ) . Since x ( n ) s x,(,~) over a single Pe-
riod (padded by N - L zeros). the original finite-duration sequence x ( n ) can be
obtained from the frequency sarnptes ( X ( 2 r k / N by means of the formula (5.1.8)
It is important to note that zero padding does not provide any additional
information about the spectrum X ( w ) of the sequence { x ( n ) ) . The L equidis'
Sec. 5.1 Frequency Domain Sampling: The Discrete Fourier Transform 401

cant samples of X(w) are sufficient to reconstruct X ( w ) using the reconstruction


formula (5.1.13). However, padding the sequence ( x ( n ) ) with N - L zeros and
computing an N-point DFT results in a "better display" of the Fourier transform
x(w).
In summary, a finite-duration sequence x ( n ) of length L [i.e., x ( n ) = 0 for
n <0 and n 2 L ] has a Fourier transform

where the upper and lower indices in the summation reflect the fact that x ( n ) = 0
outside the range 0 5 n 5 L - 1. When we sample X ( w ) at equally spaced
frequencies wk = 2 r X . / N . k = 0, 1. 2.. . . . N - 1. where N 2 L. the resultant
samples are

where for convenience. the upper index in the sum has been increased from L - 1
to N - I since x ( n ) = 0 for 11 2 L .
The relation in (5.1.18) is a formula for transforming a sequence { x ( n ) )of
length L ( N into a sequence of frequency samples (X(h)) of length N. Since
the frequency samples are obtained by evaluating the Fourier transform X ( w )
at a set of N (equally spaced) discrete frequencies. the relation in (5.1.18) is
called the discrete Fourier transform (Dm)of x ( n ) . In turn. the relation given
by (5.1.10). which allows us to recover the sequence x ( n ) from the frequency
samples

is called the inverse DFT (IDFT). Clearly, when x ( n ) has length L < N. the N-
point IDFT yields x ( n ) = O for L 5 n 5 N - 1. To summarize, the formulas for
the DFT and IDFT are

DFT

You might also like