Re: Very large FFT
- From: dbell <bellda2005@xxxxxxx>
- Date: 27 Apr 2007 09:25:36 -0700
On Apr 26, 10:01 am, "AdriaanB" <adriaan.bredek...@xxxxxxxxxxxxxxx>
wrote:
Hi all,
I am not so new to DSP but I would still like to bounce an idea off the
experts just to check that I have all my ducks in a row.
A very large FFT say 1M points is to be done on a real sequence. Am I
correct in saying that if I treat this 1mill points as a 1024*1024 matrix
and first do 1024 FFTs of length 1024 (1 FFT on each row), then transpose
the matrix, then again 1024 FFTs of size 1024(1 on each row) and then
transpose the matrix again, that I will have done my 1mill point FFT ? The
answer can be read out row by row.
Ignore the following to simplify the discussion for now:
-phase information
-complex values
Transposing the matrix is only to ease the memory usage so that I do not
have to jump around in the memory to make FFTs down the columns.
Thanks in advance.
Adriaan
_____________________________________
Do you know a company who employs DSP engineers?
Is it already listed athttp://dsprelated.com/employers.php?
Gee, that would be easy enough to do a sanity check on with a much
smaller FFT. No?
Dirk
.
- References:
- Very large FFT
- From: AdriaanB
- Very large FFT
- Prev by Date: Re: BF533 Powermode Transition
- Next by Date: Re: Transfer Function of Two Filters in Parallel
- Previous by thread: Re: Very large FFT
- Next by thread: Re: Would using the phase info makes a Noise cancelling system efficient
- Index(es):
Relevant Pages
|