[Nek5000-users] Tolerances

[nek5000-users at lists.mcs.anl.gov]

Dear Stefan,

Thank you for your quick reply!

Yes, we have already tried to refine the computational domain by rising 
the polynomial order from 7 to 9 (lx1=8 -> 10) which does not improve 
the quality of the results much (numerical artefacts in the freestream 
are damped while the numerically excited crossflow waves remain 
unchanged). 

In the case of higher resolution one disturbance wavelength would be 
resolved by 37 points on average which sounds enough to us. We will try 
to further refine the domain and to "play" with the filtering at highest 
order. Maybe we can suppress the numerical noise triggering the physical 
instability in this way...

Cheers,

David
Lars-Uve



nek5000-users at lists.mcs.anl.gov wrote:
> Do you really think 15 pressure iterations are too much? Considering you strict tolerance I would say it's pretty good! I am not sure what triggers your noise but did you try to use a high resolution?
>
> In your case TOLREL/TOLABS are not important!
>
> The tolerance you specify in DIVERGENCE is for the pressure solver. In fact if you use the PN/PN_2 scheme you'll end up with a divergence of the velocity which is comparable to DIVERGENCE. The pressure solve projects the velocity on a divergence free space. and you can control the divergence by varying the pressure tolerance (that's why we call it DIVERGENCE). In the PN/PN scheme things are different.
>
> The HELMHOLTZ controls the tolerance for all Helmholtz solves so if you don't have any passive scalars it's just for the velocity.
>
> Stefan
>