Taking The Inverse DTFT

Taking the inverse DTFT, one obtains: x[n] = 1 N x˜[n] = 1 N 1 0 X˜(f) ej2πfn df = 1 N ∞ k=−∞ X[k] 1 0
δ(f − k/N) ej2πfn df where we have interchanged the orders of summation and integration to obtain
the rightmost expression1. We note that 1 0 δ(f − k/N)ej2πfndf = ej2πkn/N 0 if 0 ≤ k ≤ N − 1
otherwise because k/N is outside of the range of integration [0, 1[ when k is outside of the interval
[0, N −1]. Therefore the inverse DFT formula is x[n] = 1 N N −1 k=0 X[k] ej2πkn/N (4.5a) Because the
signal x[n] is of finite duration, the definition of the DFT (4.1) becomes: X[k] = N −1 n=0 x[n]
e−j2πkn/N (4.5b) Formulas (4.5a) and (4.5b) constitute the DFT pair for finite-duration signals. Note
the symmetry between (4.5a) and (4.5b), the only differences being the signs of the arguments of
the complex exponentials and the 1/N factor in the inverse DFT formula (4.5a). The DFT pair (4.5)
can also be considered as a purely algebraic relation between the N numbers x[n], 0 ≤ n ≤ N − 1 and
the N numbers X[k], 0 ≤ k ≤ N − 1, with the two sets of numbers being related by a set of N linear
equations. This point of view leads to an alternate proof of the inversion formula (4.5a). Specifically,
assume that the X[k] are defined from the x[n] by (4.5b), and form the sum y[m] = = N −1 k=0
X[k]ej2πkm/N = N −1 k=0 N −1 n=0 x[n]ej2πk(m−n)/N Interchanging the order of summations
produces y[m] = N −1 n=0 x[n] N −1 k=0 ej2πk(m−n)/N Using the result that discrete complex
exponentials with period N are orthogonal over the interval [0, N − 1]: N −1 k=0 ej2πk(m−n)/N = N if
m − n is a multiple of N 0 otherwise we obtain y[m] = N −1 k=0 X[k] ej2πkm/N = N x[m] 1Although
the range of integration [0, 1[ differs from the usual range [−1 2 , 1 2 [ for inverse DTFTs, the two are
equivalent because Fourier transforms are periodic. 3 which proves the inverse DFT formula (4.5a).
This proof emphasizes that the DFT operation can be considered as a change of basis set (specifically
a rotation) in an N-dimensional vector space. In the time domain, the signal is decomposed into a
sum of orthogonal unit samples δ[n − k], while in the inverse DFT formula, the orthogonal basis
vectors are the complex exponentials ej2πkn/N . 4.1.4 Relation to discrete Fourier series We have
shown that taking N samples of the DTFT X(f) of a signal x[n] is equivalent to forming a periodic
signal ˜x[n] which is derived from x[n] by time aliasing. If the duration of x[n] is smaller than N, one
period of ˜x[n] is identical to x[n] within a factor of N. These results are the dual of those obtained in
Section 1.3 for the sampling of periodic, continuous time signals. We showed that taking N samples
per period of the periodic signal x(t) results in frequency aliasing of the Fourier series coefficients Xk.
If the bandwidth of x(t) is less than N/2T, the Fourier series coefficients of the discrete-time signal
coincide with those of the original signal x(t). Thus, there is a duality between sampling in frequency
the DTFT of a discrete-time signal to form the DFT, and sampling in time a periodic signal to form the
discrete Fourier series. In both cases, sampling produces signals that are discrete and periodic in
both the time and the frequency domain. Therefore, the discrete Fourier transform and the discrete
Fourier series are the same mathematical operation (within a factor of N). This result is to be
expected because there is a one-to-one correspondence between discrete time signals of duration N
and discrete, periodic signals with period N. Specifically, given a finite-duration signal x[n], we can
always generate a periodic signal ˜xN [n] by repeating x[n] indefinitely at intervals of N samples: x˜N
[n] = ∞ r=−∞ x[n + rN] = x[n] ∗ ∞ r=−∞ δ[n − rN] (4.6a) Conversely, given a periodic signal ˜xN [n],
we can form a finite-duration signal x[n] by multiplication with a rectangular pulse of length N:
x[n]=˜xN [n]RN [n] = x˜N [n] if 0 ≤ n ≤ N − 1 0 otherwise (4.6b) The Fourier series coefficients X˜k of
the periodic signal are related to the DFT of the finiteduration signal by the formula: X˜k = 1 N X[k]
(4.7) To summarize, computing the N-point DFT of a signal implicitly introduces a periodic signal with
period N, so that all operations involving the DFT are really operations on periodic signals. These
operations will give the same results as operations on finite-duration signals providing that the
durations of all the signals involved in these operations are less than N

Taking The Inverse DTFT

Uploaded by

Copyright:

Available Formats

Taking The Inverse DTFT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Taking The Inverse DTFT

Uploaded by

Copyright:

Available Formats

Taking the inverse DTFT, one obtains: x[n] = 1 N x˜[n] = 1 N 1 0 X˜(f) ej2πfn df = 1 N ∞ k=−∞ X[k] 1 0

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.