Shake Error with Ewald but not PPPM

​Sorry Steve,

I forgot to cc the mailing list on the last email.

You say it happens when you switch to NPT. Is your box volume
changing rapidly, or is the change large? If the former,
then that can lead to bad dynamics. If the latter, then you
may need to break your run into smaller runs, so that the
KSpace solver can choose different grid spacings (or # of
KSpace vectors in the case of Ewald) to give you the
desired accuracy. E.g. you can’t double the volume and
use the original grid size and get the accuracy you specified.