Re: FFTW: a problem with in-place 2D r2c transform
- From: "young" <anna.krechtchouk@xxxxxxx>
- Date: Wed, 15 Mar 2006 03:26:11 -0600
Thanks a lot for your kind reply, Steven. I feel very flattered to be
answered by the author himself, and even more confused to continue asking
my dumb questions, but:
1) I read this part of the manual. I didn't realize that allocating more
than is needed was a mistake.
2) I've changed the array sizes to the exact values in your reply. I still
get zero-frequency of 351 for in-place, and 432 for out-of-place.
What am I doing wrong??
Thanks again
The array format for r2c transforms is described in detail at:
http://www.fftw.org/doc/Multi_002dDimensional-DFTs-of-Real-Data.html
In particular, note that
1) the output is n[0] * (n[1]/2 + 1) in size, not n[0]*n[1]
2) the input is n[0] * n[1] in size for out-of-place and n[0] * 2 *
(n[1]/2+1) in size for in-place, where the latter reflects padding to
make room for the complex output.
This is all described in the manual.
Cordially,
Steven G. Johnson
.
- Follow-Ups:
- Re: FFTW: a problem with in-place 2D r2c transform
- From: stevenj
- Re: FFTW: a problem with in-place 2D r2c transform
- References:
- FFTW: a problem with in-place 2D r2c transform
- From: young
- Re: FFTW: a problem with in-place 2D r2c transform
- From: stevenj
- FFTW: a problem with in-place 2D r2c transform
- Prev by Date: Re: Possible to predict phase in a spectrogram?
- Next by Date: Re: frequency shifting (compressing?)
- Previous by thread: Re: OT Steven Johnson
- Next by thread: Re: FFTW: a problem with in-place 2D r2c transform
- Index(es):
Relevant Pages
|
Loading