[SciPy-user] fft-problems

[Lev Givon]

Received from Alpha on Sun, May 21, 2006 at 03:36:12PM EDT:
> Hello everybody,
> 
> playing around with scipy I'm stuck with a problem with fftconvolve
> in scipy.signal and other fft-related stuff. Fft operations yield too
> many zeros, much to high coefficients and many NANs in the matrix.
> 
> The code below illustrates my problem. I'd be thankful for any help
> or comments. 
> 
> Thanks, 
> Alpha
> 
> 
> 
> import scipy
> from scipy import *
> a =  array([[1,1,1],[1,1,1],[1,1,1]])
> print(a)
> print(scipy.signal.convolve(a,a))
> print(scipy.signal.fftconvolve(a,a))
> # both results should be equal or at leas similar, yet they aren't
> 
> 
> This is the output on my system:
> >>> print(a)
> [[1 1 1]
>  [1 1 1]
>  [1 1 1]]
> 
> >>> print(scipy.signal.convolve(a,a))
> [[1 2 3 2 1]
>  [2 4 6 4 2]
>  [3 6 9 6 3]
>  [2 4 6 4 2]
>  [1 2 3 2 1]]
> 
> >>> print(scipy.signal.fftconvolve(a,a))
> [[  2.54728497e+014   2.40207507e+287   1.09183674e+277
> 1.98812999e+286 1.42688102e+286]
>  [  8.09194231e+284   4.34718284e+284  -6.99555293e+281
> 6.99555647e+281 -4.34718284e+284]
>  [  6.80511448e+286  -1.42688102e+286  -1.98812999e+286
> 0.00000000e+000 0.00000000e+000]
>  [  0.00000000e+000   0.00000000e+000   0.00000000e+000
> 0.00000000e+000 0.00000000e+000]
>  [  0.00000000e+000   0.00000000e+000   0.00000000e+000
> 0.00000000e+000 0.00000000e+000]]
> 
> -- 
> 
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-user

convolve and fftconvolve returned consistent answers when I tried the
above code on scipy 0.4.8 built with numpy 0.9.8 and fftw 3.0.1.
 
						L.G.