#58 Damping not applied in ghost cells

[Anonymous]

Originally created by: Bourdin.KIS (code.google.com)
Originally owned by: Bourdin.KIS (code.google.com)

If you look in hydro.f90/subroutine udamping, you will find, that damping is only applied to the physical cells, not the ghost cells.

If one has a boundary condition, which tries to have a smooth continuation in the ghost cells, this means, that after damping (especially if damping is strong) the continuation in the ghost cells is nomore smooth and strong gradients can arise at the boundary.

I don't know if this is maybe even an intended feature for some of you, so I just want to write this down and ask, if someone can review this issue, please.

Note, that 'udamping' is being called relatively late in subroutine duu_dt (short before diagnostics), but relatively early in the mn_loop in equ.f90/subroutine pde.

So, as far as I can judge, the non-smooth boundaries can have undesirable consequences - see attachment image, which are wiggels I can't remove by applying more damping, nor with stronger diffusion or hyper3 diffusion/viscosity parameters. - Otherwise, this particular simulation is in a very good hydrostatic equilibrium state.

[Anonymous]

Originally posted by: Bourdin.KIS (code.google.com)

(No comment was entered for this change.)

Labels: -Type-Review Type-Defect

[Anonymous]

Originally posted by: wdobler@gmail.com

Are you sure that applying udamp to the ghost cells would help? I think it would not make a difference at all.

We
(1) apply the boundary conditions to set the ghost cell values,
(2) then calculate all required derivatives,
(3) and apply velocity damping.

An extra item
(4) apply damping to the ghost cells (or do /anything/ to the ghost cells)
will not have any effects, because before calculating derivatives for the next time, we apply the boundary conditions again and thus erase any information that may have been in the ghost cells.

[Anonymous]

Originally posted by: Bourdin.KIS (code.google.com)

I'm not sure, in how far the values from the ghost cells are relevant/being used _after_ the velocity damping and before the next mn_loop. After the udamping call I just see a lot of relevant code - of which I don't know large parts. Therefor this request for comments.

True, the derivatives are stored in the df-array before the udamping call.
So you say, I can be sure that even if I would completely destroy the ghost cells after the damping, it should not have any effect? ...this could be tested.