[gmx-users] Timesteps don't match

[Mark Abraham]

On Fri, Mar 15, 2013 at 7:10 PM, Alexander Peyser <peyser.alex at gmail.com>wrote:

> Dear all,
>
> I'm receiving a "timesteps don't match" error, due to a wobble in the time
> between steps for my xtc trajectory file. The source of the wobble is
> mysterious.
>
> I'm running gromacs-4.5.5 with plumed patches for a very long trajectory
> using double precision code with a time step of 2fs and xtc frames every
> 200fs. The trajectory was run in 24 hour chunks, with a continuation with
> -append over all the chunks. The timing fit precisely the values set over
> many, many chunks.
>
> Unfortunately, I was trying to consolidate some disk space and lost my log
> file, which kept me from simply running mdrun_d with -append to continue
> the trajectory as I had been doing. I started a new trajectory with the
> checkpoint file by extending the tpr with tpbconv, and then continued
> appending to the output from that. The beginning time is correct, beginning
> right after the end of the first trajectory. However, every frame is
> separated from the next by 187.5 fs or 218.75 fs, repeatedly. I've checked
> my input files, my cpt checkpoint and my tpr run file --- there are no step
> values like that at all, and the xtc output time continues to be set as
> every 100 steps, with a step time of 2fs.
>
> Where should I look to identify the source of this wobble? Has anyone seen
> anything similar?
>

Sounds pretty weird. I haven't got any theories on how that could happen.
If you've got input files that can reproduce the issue, please file it on
redmine.gromacs.org.

Mark