[FMDenaro]

Originally Posted by HectorRedal

I have kept constant the dt/h ratio between grids.

however be careful that, depending on your discretization, also the diffusive part enters into the stability constraints

[HectorRedal]

For stability condition I take into account two terms: One term for the convective stability condtion (which should be less than h/u) and another term for the diffusive stability condition (which should be less than h^2/(2 nu), being nu the kinematic viscosity)

[FMDenaro]

Originally Posted by HectorRedal

For stability condition I take into account two terms: One term for the convective stability condtion (which should be less than h/u) and another term for the diffusive stability condition (which should be less than h^2/(2 nu), being nu the kinematic viscosity)

actually, both terms act simultaneously giving a CFL(Re_h) stability region. It can happens that at low Re_h you cannot take dt/h constant but you have to reduce further the time-step.

Thus, I think you have first to check if oscillations still appear when you solve, but on the same grid of value h, with the two time steps

[HectorRedal]

I have already checked this point: us9ing the same grid but varying the dt. When I use a coarse grid, and I reduce the dt, the small oscillations does not appear. This problem (small oscilations) only appear when using a fine grid and reducing the dt.

[FMDenaro]

Originally Posted by HectorRedal

I have already checked this point: us9ing the same grid but varying the dt. When I use a coarse grid, and I reduce the dt, the small oscillations does not appear. This problem (small oscilations) only appear when using a fine grid and reducing the dt.

that probabily means you are approaching the limit of the stability region and on that fine grid (small Re_h) you need a smaller cfl (smaller dt).

[HectorRedal]

I have reduced the threshold of the iterative solver, along with the time step, and the problem has been eliminated. There is not any oscilation in the solution as the time step is reduced.

Thanks you all for your kind support and help.

[FMDenaro]

Originally Posted by HectorRedal

I have reduced the threshold of the iterative solver, along with the time step, and the problem has been eliminated. There is not any oscilation in the solution as the time step is reduced.

Thanks you all for your kind support and help.

yes, at small dt you need to improve the convergence threshold otherwise we get often small oscillation from a time step to an other