handling periodic images

[Noam Bernstein]

Hi - I'm running a simulation of a molecule with periodic boundary
conditions, and I get rather large discrepancies between finite
difference approximations of the force and the analytical forces.
For a big unit cell (molecule is about 5x5x5 A, cell is 25x40x40),
these discrepancies happen only for a few atoms, in a few cartesian
directions.  In the process of trying to make a smaller test case, I
found  that the discrepancies became much larger for a smaller unit
cell, and affected all atoms and all directions.

The only thing I can think of is something about how periodic images
are handled (i.e. not just minimum image convention).  However,
I've scoured the mailing list and documentation, but I can't find any
mention of parameters that might affect the number of periodic
images that are taken into account in various calculations
(matrix elements, Ewald, maybe other things).  Is there something
that I'm missing?

                                                          Noam