I found where is the problem.
I modified recently the '.afnirc' file in this way:
AFNI_COMPRESSOR = BZIP2
so that AFNI brik files were automatically bzip2 compressed (es.: DIFF+orig.BRIK.bz2).
If AFNI compute DWI analysis
(3dDWItoDT -prefix eig -automask -eigs $N DIFF+orig
ecc.) with this kind of file,
I get no good data (FA, main diffusion eigenvectors maps, ecc. incorrect). If I disable compressed saving and retry
the same DWI analysis, It's all ok.
Do you ever want that I send you my data?
Thanks for support.
Ciao
Giovanni