Hi there, just to follow up:
This was inspired by our desire to replicate the (apparent) interpolation quality afforded by FLIRT -sinc. It appears that there remain subtle differences between the two programs with respect to the window type used. I have better success with 3dAllineate in general as a registration tool, so I was hoping to elucidate what would cause the appearance of more Gibbs ringing using wsinc5. I now believe that FLIRT's apparent higher quality output is simply due to a less accurate window function.
I notice subtle improvements by expanding the sinc window size to about 7 or 9 voxels. I generally use 7 now. This is the same as the default parameter used by FLIRT.
The window functions between the two programs are different. FSL uses a blackman window (default), while AFNI implements the exact blackman window (default). I'm not sure about this, but I think that FSL's window choice decreases the appearance of Gibbs ringing because the window smooths over details more than the exact blackman implemented in AFNI. It is very obvious that, while introducing more ringing, wsinc5 also preserves more underlying detail
I can email you (rwcox) some photos if you would like, I have no where to host them currently.
Thanks,
Joe