dear alex
If you specify reading of box parameters from the input coordinate file,
(they should be the last line of that file), these will replace box X etc
in the mdin file. The NFFT1-3 don't have to be bigger than corresponding
box parms. With your other params, use somewhere near a 1 angstrom
density, i.e. NFFT1 close to BOX X etc. 10% more or less won't matter
much. By the way, pick the nfft values to be multiples of powers of 2,3,5
for speed in fast fourier transform of PME. For example, you use 37, 42
and 51. Instead use 36 (=2x2x3x3), 40 (=5x2x2x2) and 50 (=5x5x2) and
things will go much faster at essentially the same accuracy. This stuff
is a pain to remember, so in amber 6 it is (hopefully) much more
automatic.
hope this helps
Tom Darden
On Thu, 9 Mar 2000, Alexander Issanin wrote:
>
> Dear Amber users!
>
> I have a question about PME parameters specified in the end of sander
> input (.mdin) file e.g. PME box parameters, size of charge grid and so
> on. In AMBER 5 manual it is written that "this information must be
> specified and _overrides_ the box information specified in the parm
> file". On the other hand when I start NVT-dynamics after NPT, I can find
> in output file something like:
>
> ----------------------------------------------------------
> EWALD SPECIFIC INPUT:
>
> Box X = 36.723 Box Y = 42.195 Box Z = 50.815
> Alpha = 90.000 Beta = 90.000 Gamma = 90.000
> NFFT1 = 37 NFFT2 = 42 NFFT3 = 51
> Cutoff= 9.500 Tol =0.100E-04
> Interpolation order = 4
> Largest sphere to fit in unit cell has radius = 18.361
> Ewald Coefficient = 0.29051
> -----------------------------------------------------------
> (parameters from .mdin file)
> and then
>
> -----------------------------------------------------------
> Largest sphere to fit in unit cell has radius = 16.669
> NEW EWALD BOX PARAMETERS from inpcrd file:
> A = 34.82838 B = 33.33781 C = 45.31798
>
> ALPHA = 90.00000 BETA = 90.00000 GAMMA = 90.00000
> -----------------------------------------------------------
>
> My questions are:
> 1) Should I put new actual box parameters in .mdin file after
> NPT-dynamics run (or they will be read automatically from .inpcrd file)?
> 2) How important are NFFT1..NFFT3 parameters for the accuracy of the
> run? Is it better, when BOXX/NFFT > 1.0 or < 1.0?
> 3) What meaning has "Ewald error estimate" in the output file? What
> value supposed to be appropriate?
>
> Thanks for the help!
>
> Alex Issanin.
>
Received on Thu Mar 09 2000 - 11:28:30 PST
