Re: zero padding radix - 2 FFT

forums_mp@xxxxxxxxxxx wrote:
For starters, I'm still a newbie muddling my way through the various
texts and fourier terminologies. I'm of the impression that given an
FFT with an odd size. Zero padding the FFT to the next power of two
amounts will do the trick. Barring efficiency reasons with regards to
odd sizes, I'm only interested in how the padded and non padded version
could be equivalent. Now utilizing matlab as my analysis tool.
Given:

x=[0,1,2,3,4]
x =
0 1 2 3 4
y=fft(x)'

y =
10.0000
-2.5000 - 3.4410i
-2.5000 - 0.8123i
-2.5000 + 0.8123i
-2.5000 + 3.4410i

Now zero pad to the next power of two.

x=[0,1,2,3,4,0,0,0]
x =
0 1 2 3 4 0 0 0
y=fft(x)'

y =

10.0000
-5.4142 + 4.8284i
2.0000 - 2.0000i
-2.5858 + 0.8284i
2.0000
-2.5858 - 0.8284i
2.0000 + 2.0000i
-5.4142 - 4.8284i


How is the result from the result from the padded and non-padded
version equivalent or is there my understanding flawed?

They are both sets of points on (samples of) the same underlying continuous
complex curve, just spaced differently. The longer you pad your sequence
the closer the sample points get to each other in the frequency dimension.


Bob
--

"Things should be described as simply as possible, but no simpler."

A. Einstein
.

Relevant Pages

  • Re: Zero Padding in radix 2 FFT
    ... When I am provided with samples which are not of power of 2 and I need ... I do FFT using Cooley Tukey algo and observe that the frequency ... Does it mean zero padding is not after all a good idea if you have ... Is it a good idea to resample and then apply or use a different FFT ...
    (comp.dsp)
  • Zero Padding in radix 2 FFT
    ... When I am provided with samples which are not of power of 2 and I need ... I do FFT using Cooley Tukey algo and observe that the frequency ... Does it mean zero padding is not after all a good idea if you have ...
    (comp.dsp)
  • zero padding radix - 2 FFT
    ... FFT with an odd size. ... Zero padding the FFT to the next power of two ... Now zero pad to the next power of two. ...
    (comp.dsp)
  • Re: Phase/twiddle factors calculations
    ... I will consider the final stage of a decimation-in-time FFT where two N/2-length FFTs are combined to form one N-length FFT. ... One N/2-length FFT will have been been generated from the 'even' time-domain samples and the other from the 'odd' samples. ... When the two half-length FFTs are combined to form a single full-length FFT, an increasing phase-adjustment must be made to the frequency components from the 'odd' FFT before summing them with their 'even' counterparts. ... Conceptually, two 'sweeps' through the N/2-length FFTs are necessary to generate N output components, but the FFT Butterfly applies some cunning trigonometry so that only one sweep is needed and that only N/2 complex phase-adjustment factors are needed. ...
    (comp.dsp)
  • Re: How to interpolation in frequency domain will not affect time domain signal?
    ... a N-point spectrum was obtained after FFT to ... But the sampling frequency is not big enough ... I try to interpolate in the N-point spectrum to ... you may be zero padding incorrectly. ...
    (comp.dsp)
Loading