Could we calculate FFT/IFFT for N samples based on N/2 samples?

On Apr 9, 1:54 pm, "Rune Allnor" <all...@tele.ntnu.no> wrote:
> On 9 Apr, 17:05, "karac...@gmail.com" <karac...@gmail.com> wrote:
>
> > Unfortunately, the author of the library might forget to leave any
> > document about their work. The problem is "I need to calculate FFT for
> > 2048 samples while this library supports only 1024 FFT calculation".
>
> Ah. So you use in-house code where the maximum FFT size is
> hard-coded...
>
> ...
>
> > It is possible for me to use other library which supports full-
> > functions of FFT calculation. However, I think usage of that kind of
> > library will break down my effort of understanding FFT.
>
> > Could you help me to understand more about my friend's answer?
> > Especially what is "weight" parameter? How to find the way of
> > rearranging?- Skjul sitert tekst -
>
> Depending on how much time you have available, either get
> another FFT library or implement it yourself, with sufficient
> flexibility to at least handle an FFT length of any power of 2.
> There are plenty of FFT codes publicly available which you
> can use as template (see e.g. "Numerical Recipies in C").
>
> Proceeding with the current library is a waste of time and
> will cause a lot more problems than it is worth.
>
> Rune

Hi Rune,

> > Unfortunately, the author of the library might forget to leave any
> > document about their work. The problem is "I need to calculate FFT for
> > 2048 samples while this library supports only 1024 FFT calculation".
>
> Ah. So you use in-house code where the maximum FFT size is
> hard-coded...
Exactly, this is why I have problems


> Depending on how much time you have available, either get
> another FFT library or implement it yourself, with sufficient
> flexibility to at least handle an FFT length of any power of 2.
> There are plenty of FFT codes publicly available which you
> can use as template (see e.g. "Numerical Recipies in C").
>
> Proceeding with the current library is a waste of time and
> will cause a lot more problems than it is worth.
It seems that implement FFT is easier than I expected. Even though my
code does not the best, it might be flexible to handle my purpose.

You and Fred gave me the honest advices. Thank you very much. It is
worth for me to post my question in this group.