[gmx-users] problems about number of frames in the edr file

[Mark Abraham]

On 21/07/2011 3:04 AM, Zhuyi Xue wrote:
> Hi, there,
>
> I have a edr file that looks containing about 2000 frames based on its 
> size as well as the result from gmxcheck:
>
> frame: 75151040 (index      0), t: 150302.094
> Reading energy frame      2 time 150320.000
> Timesteps at t=150310 don't match (7.90625, 10)
> Reading energy frame   2000 time 170300.016
> Timesteps at t=172690 don't match (10, 5.64062)
> Last energy frame read 2240 time 172695.656
>
> Found 2241 frames.
>
> But when I use g_energy to get the temperature of each frame in this 
> edr file, only 4 frames could be extracted, the last 10 lines of the 
> result xvg file is as following:
>
> @ legend on
> @ legend box on
> @ legend loctype view
> @ legend 0.78, 0.8
> @ legend length 2
> @ s0 legend "Temperature"
>   150302.093750  301.350067
>   150310.000000  299.902710
>   150320.000000  303.125641
>   150330.000000  300.704010
>
> I have tried eneconv to fix it, but no luck. How to could fix it and 
> get the values for all frames?

The energy file format contains various quantities related to the 
running average and running squared deviation. These require the time 
step to be the same between each frame. As such, the analysis tools 
expect the time step not to change. Some tools get confused if it does 
change. eneconv -settime allows you to artificially manipulate the 
times, if you think you have a good reason for concatenating the 
energies from dislocated chunks of a trajectory. As soon as you do so, 
the averages and fluctuations reported by g_energy are wrong.

Doing restarts with mdrun -append in the manner suggested on the GROMACS 
webpage will avoid needing to deal with any of this stuff.

Mark

> Another question is what does it mean that the timesteps at t=... 
> don't match? Will it affect the result of eneconv when connecting 
> multiple edr files?
>
> Thank you
> Zhuyi