[gmx-users] Re: Questions WHAM

[lloyd riggs]

-------- Original-Nachricht --------
> Datum: Wed, 18 Apr 2012 08:18:27 -0400
> Von: "Justin A. Lemkul" <jalemkul at vt.edu>
> An: Discussion list for GROMACS users <gmx-users at gromacs.org>
> Betreff: Re: [gmx-users] Re: Questions WHAM

> 
> 
> lloyd riggs wrote:
> > -------- Original-Nachricht --------
> >> Datum: Wed, 18 Apr 2012 03:13:09 +1000 Von: Mark Abraham
> >> <Mark.Abraham at anu.edu.au> An: Discussion list for GROMACS users
> >> <gmx-users at gromacs.org> Betreff: Re: [gmx-users] Re: Questions
> > 
> >> On 18/04/2012 12:10 AM, lloyd riggs wrote:
> >>> Included below (although Im terrified youll invalidate massive
> parallel
> >> work with a single comment) is the .mdp file I used for 10 of the runs,
> >> basically it was the only way I could get it to run. The time step is
> small
> >>  because of the small molecule involved, however I painstakingly looked
> up
> >> all parameters ten times.  Basically its only violation was with links,
> >> which seems to think S-O---H isnt supposed to roatate so much, but the
> >> molecules degrees of freedom (1/2 can rotate 360 from the other along a
> >> central axis) etc... PR p-coupling is poor for stability during
> >> equilibration (since gen_vel = yes), and it's always poor to start your
> >> production run (i.e. your pulling) without prior equilibration.
> Equilibrate
> >> with berendsen, then switch to PR for a bit and only then worry about
> >> pulling.
> > 
> > I had woundered as I always have to throw out the first 2 picosecounds
> about
> > the init velocity.  I did equilibrate each first untill there was no
> change
> > with bereendsen and nose-hoover, so from my mistake thats what I have
> had to
> > do for the data.  Is there any other critiques, as I COULD USE THEM.
> > 


> 
> I would say that if you are regenerating velocities, you need to get rid
> of a 
> lot more than 2 ps, especially with Nose-Hoover.  Even Berendsen will
> likely not 
> re-converge that quickly.

It only gave me a problem beyond 2 ps for 1 of 10 runs, but I wanted to stay uniform.  In the one I had to cut out the first 6 picosecounds. Mostly though they had been equed for a few hundred pico secound (until no change) and my time step was extreemly small, so the velocities I think are only applied very short on my scale here...but I do not know if thats why they EQ fast.

Yeah trying to get all the columns uniform though can be a pain, I like emacs scripts now.


Stephan
-- 
NEU: FreePhone 3-fach-Flat mit kostenlosem Smartphone!                                  
Jetzt informieren: http://mobile.1und1.de/?ac=OM.PW.PW003K20328T7073a