Small timestep make the simulation not converge
 

Hi

I was stuck by the convergence problem for whole the last week. Today I accidentally find out why. The reason is I used too small time step. But the FAQ suggests us to use small timestep to make our simulation converge. Could you explain why sometimes, to the contrary of the theory, small time step might even prevent convergence?

If your time step gets too small then round off errors can become large. Then the noise from round off error means convergence is difficult.