[Scipy-tickets] [SciPy] #1232: Support fft normalization once added to numpy

SciPy Trac scipy-tickets@scipy....
Thu Jul 15 08:25:08 CDT 2010


#1232: Support fft normalization once added to numpy
-----------------------------------------------+----------------------------
 Reporter:  dgoldsmith                         |       Owner:  somebody
     Type:  enhancement                        |      Status:  new     
 Priority:  normal                             |   Milestone:  0.9.0   
Component:  Other                              |     Version:  0.7.0   
 Keywords:  fft normalization keyword unitary  |  
-----------------------------------------------+----------------------------

Comment(by charris):

 Yes, the pair needs to be normalized, but the divide by N normalization is
 often not the right one. I usually normalize the transforms by the spacing
 of the samples, seconds or centimeters in the forward case, hz or 1/cm in
 the reverse case. This deals properly with resampling and gives everything
 useful units. The usual normalization corresponds to taking the spacing of
 the original data set to be 1 and doesn't work properly if one wants to
 interpolate the original data by extending the transform with zeros. This
 normalization also preserves the sum of squares if one multiplies the sums
 by the respective sample spacings so they look like integrals. So I don't
 want numpy doing my normalization, numpy gets it wrong for many cases and
 it is a hassle to undo the damage. Better would  be to add a keyword for a
 scaling factor. FFTW got this aspect right.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1232#comment:10>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list