Re: FFTW versus NRC FFT
- From: Bruce Scott TOK <Use-Author-Supplied-Address-Header@[127.1]>
- Date: Fri, 5 Jan 2007 12:57:07 +0100 (MET)
As a postscript to Steven Johnson's detailed messages, simply remember
that the complex-to-real and real-to-complex Fourier transforms are
critically phase dependent, and to know in detail what each package
assumes about these phases.
I ran into this while using FFTW on my grid -\pi < y < \pi, where the
branch cut falls between grid nodes. FFTW assumes you are on 0 < y < 2\pi
and that the first grid node is on the branch cut.
--
ciao,
Bruce
drift wave turbulence: http://www.rzg.mpg.de/~bds/
.
- Follow-Ups:
- Re: FFTW versus NRC FFT
- From: stevenj
- Re: FFTW versus NRC FFT
- From: Blacky
- Re: FFTW versus NRC FFT
- References:
- FFTW versus NRC FFT
- From: Blacky
- Re: FFTW versus NRC FFT
- From: stevenj
- Re: FFTW versus NRC FFT
- From: stevenj
- FFTW versus NRC FFT
- Prev by Date: Re: poisson equation with periodic boundary conditions using finite difference
- Next by Date: Re: FFTW versus NRC FFT
- Previous by thread: Re: FFTW versus NRC FFT
- Next by thread: Re: FFTW versus NRC FFT
- Index(es):