[phenixbb] lower limits on grid spacing/binning

Ralf W. Grosse-Kunstleve rwgk at cci.lbl.gov
Tue Oct 30 10:39:54 PDT 2007

Probably, the original poster already got more than he bargained
for, but for the records, a few more comments:

>    All FFT based structure factor programs require that the sampling
> rates along each axis be even.

I don't think this is true, unless I'm misunderstanding what "sampling
rate" means. The FFT in phenix is based on FFTPACK (written in the
80s) which works for any number of grid points. FFTw also supports
arbitrary gridding.

> They may have other required factors
> depending on the space group, but they will be happy to inform you
> if you make a choice it doesn't like.  They are also more efficient
> when the prime factors of the sampling rates are small numbers.  Try
> to stick with multiples of 2,3, and 5 if possible.

That's good advice. FFTPACK is fastest for transform sizes that are
multiples of 2,3,4 and 5. The map calculation algorithms in phenix
automatically take this into account.

> All FFT programs will fail if you
> sample your map courser than twice that frequency, as SFALL did for
> you.

This may be true for SFALL, but not for the FFT algorithms in the
phenix libraries ($PHENIX/cctbx/include/cctbx/maptbx/structure_factors.h).
The critical reference is:

  David A. Langs (2002), J. Appl. Cryst. 35, 505.

As short as it is, this was an incredibly important paper. In
retrospect, it is amazing that it took so long for someone to discover
the trick.


More information about the phenixbb mailing list