| Author |
 Topic  |
|
|
Mike Buess
USA
3037 Posts |
 Posted - 03/19/2005 : 11:50:29 AM
|
Origin Version (Select Help-->About Origin): 7.5
Operating System: XP
In pulsed NMR it is customary to divide the values of the first complex point by two prior to FFT. Does anyone know if Origin C's FFT and FFT2 functions do this? I know that they are based on the NAG FFT functions but I can find no discussion about this issue in c06ecc_cl05.pdf (nag_fft_complex).
Mike Buess
Origin WebRing Member |
|
|
easwar
USA
1965 Posts |
Posted - 03/21/2005 : 12:35:31 PM
|
Hi Mike,
We do not perform any scaling such as you mention here, prior to calling the NAG functions. I have passed on your query to NAG and will post a reply once I hear from them.
Easwar
OriginLab
Edited by - easwar on 03/21/2005 12:36:20 PM |
 |
|
|
easwar
USA
1965 Posts |
Posted - 03/21/2005 : 1:07:04 PM
|
Hi Mike,
The FFT, IFFT etc routines call the nag_fft_multiple functions from the library. NAG replied saying that these functions scale the entire data by 1/SQRT(N) on both forward and inverse. They do not handle the 1st data point in any special way. NAG further comments:
" What the customer might be referring to is that, in some (non-NAG) implementations of real-to-complex FFTs and their inverses, the user indeed must manipulate the mean term of the forward transform -- in such implementations, the forward transform returns a value for the mean term that it not immediately usable by the inverse algorithm, requiring a manipulation by the user. Be assured our implementation is consistent with the mathematical definition of the transform, and no strange manipulations like this are required"
Easwar
OriginLab
|
|
|
|
Mike Buess
USA
3037 Posts |
Posted - 03/21/2005 : 4:36:40 PM
|
Thanks, Easwar. I've been using such manipulation code and wanted to be sure it wasn't redundant. Looks like it might not be necessary.
Mike Buess
Origin WebRing Member |
|
|
| |
Topic |
|
Initial time point in FFT