Temperature trouble

[Martín Lema <mlema "at@at" unq.edu.ar>]

 Dear CCLs:
In my MD simulations, I start heating up the system using
 Berendsen´s
 algorithm and when it reaches 300 K I switch to Andersen´s. The
 thing is
 the average temperature of the system then stabilizes around 294 K.
The system is a myoglobin of 2134 atoms (including hydrogens),
 implicitly
 solvated using the SASA algorithm and using the CHARMM27 parameter set.
I have tried different collision frequencies for the thermostat,
 longer
 times, longer cutoffs, etc, but I can´t find the origin of this
 discrepancy.
Collision values of 0 (which means no application of the thermostat)
 and
 166 (which means velocity is reassigned for every atom in every step)
 renders the system at 300 K, but it is not the case for more reasonable
 values in the range 0.02 - 5.
 The trouble persist when performing the simulation in vacuo or using
 other
 solvation algorithms.
Does someone figures out where the trouble is? clues are welcome at
 mlema "at@at" unq.edu.ar. Useful responses will be summarized.
 Martin Lema
 National University of Quilmes
 Argentina
.