error block averaging technique

[Aigul Khakimova]

Dear plumed users and developers,

I’ve trying to analyse the convergence of WT metadynamics and calculate error using the block averaging technique, but I have strange and non-physical results, that I can't understand.

System: protein (~5200 atoms) + ligand (70 atoms)

Simulation time: 100 ns

FES of the system:

How I count the error:

1) reweight the system using plumed
2) find maximum bias
3) count weighted collective variables
4) find minimum and maximum in file with reweighted collective variable
5) find FES for each block (use script https://github.com/plumed/masterclass-21-4/blob/main/data/do_block_fes.py) from plumed tutorial 
6) count error for each block

I got the error:

 

Sometimes for another simulation, I got the error:

I have realised, that my system doesn't converge. That the error decrease with the block size it is not physical. Our values must become more independent with increasing block size -> error increase. But in my case, the error decreases or changes very fast... maybe something wrong, but what is wrong? Or is it doesn't matter, matter only that the error changes during all time and we need to find another CV or decrease simulation time?

Thank you!

Best regards, 

Aigul