Re: [OctDev] FFT normalization

[JuanPi <aj...@gm...>]

Hi Peter,

Thank you very much. I think this should be somewhere in the manual
and the documentation. The scaling factor of the F. transform and its
inverse differs a lot from discipline to discipline, so it is better
to always state it explicitly.


JPi


On Tue, Sep 21, 2010 at 12:39 PM, Peter L. Soendergaard
<pe...@so...> wrote:
> tir, 21 09 2010 kl. 10:31 +0200, skrev JuanPi:
>> Hi everybody,
>>
>> According to FFTW
>> http://www.fftw.org/faq/section3.html#whyscaled
>>
>> the fft is not normalized...hence ifft(fft(y)) = length(y)*y
> To clear things up:
>
> The fft is not normalized, but the ifft is normalized in such a way that
> ifft(fft(y)) == y
>
> In FFTW, ifft is not normalized this way, so ifft in FFTW and
> Octave/Matlab differs, but the fft's are the same.
>
> Normally, if one talks about a normalized fft, the normalization is
> fft(y)/sqrt(length(y)). This gives fft and ifft the same normalization
>
> /Peter
>
>



-- 
JuanPi Carbajal
-----
"I have come to believe that the whole world is an enigma, a harmless enigma
that is made terrible by our own mad attempt to interpret it as though it had
an underlying truth."
-- Umberto Eco
-----
www.ailab.ch/carbajal