Re: Very large FFT


.... a more mathematical answer to that please ?

It is clear that when the recipe says you add salt and two eggs it is best
to add exactly that. I would just rather like to know the why without
having to follow it blindly.

In the jjj.de .pdf file on page 329 and some pages before where Jörg shows
a Image processing 2D FFT, the difference is exactly the extra Twiddle
factor you have to make to get back to a 1D 1M point FFT.

R




Yes it is clear to me that this is the same Twiddle array that I
previous
calculated that for increasing the speed of the FFT calc, but why is
he
making it then again ?

It looks like
- FFT with twiddle
-Twiddle again (???)
- FFT with twiddle
-final answer

I noticed the same in a PhD thesis on radio astronomy in the appendix
C
where the author made an extra Twiddle step. In the referenced
presentation from Dillon Engineering there was also this step.

What is the purpose of it? Is the 1D FFT calculated as 2D matrix
really
that much different to the image processing 2D FFT ?


Without it you don't get a 1M FFT when you are done. It would be nice
not to have to do it, but if you don't do the extra twiddle multiply
you would have a 2D FFT 1k x 1k FFT and not a 1M FFT.

Tom





_____________________________________
Do you know a company who employs DSP engineers?
Is it already listed at http://dsprelated.com/employers.php ?
.

Relevant Pages

  • Re: Very large FFT
    ... Yes it is clear to me that this is the same Twiddle array that I previous ... FFT with twiddle ... where the author made an extra Twiddle step. ... presentation from Dillon Engineering there was also this step. ...
    (comp.dsp)
  • Re: Very large FFT
    ... transpose and repeat the rows again and then ... Why is the extra twiddle factor ... I wondered some time ago about an FFT of a whole CD, ...
    (comp.dsp)
  • Re: Very large FFT
    ... calculated that for increasing the speed of the FFT calc, ... FFT with twiddle ... where the author made an extra Twiddle step. ...
    (comp.dsp)