[Nek5000-users] time derivative

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

Hi Francesco,

I would use a higher-order derivative, e.g.,

(3u^n - 4u^n-1 + u^n-2)/ 2dt would be 2nd order accurate.

You could also use centered differences,  (u^n+1 - u^n-1) / 2dt

to compute du/dt at t^n.

The best way to verify your results is to take lim dt--> 0 and
see that you realize the expected convergence.

There should be no difficulty with the characteristics scheme,
as far as I can tell.

Paul

On Tue, 14 Dec 2010, nek5000-users at lists.mcs.anl.gov wrote:

> thank you.
>
> i do have a follow up question.
> My case is a DNS of a turbulent channel.
> I was hoping to use the the characteristics method (IFCHAR=T.)
> and a target CFL number of 1.5 - 2.0 .
>
> To compute time derivatives I am using a normal finite difference 
> [u(t)-u(t-dt)]/dt
> but the results are suspicious.
>
> Given the large time step, do you think it is possible to
> Have reliable accelerations with this method?
> Are there any better ways to compute them?
>
> thank you a lot.
> francesco
>
>
>
>
> On 12/13/2010 04:59 PM, nek5000-users at lists.mcs.anl.gov wrote:
>> 
>> Hi Francesco,
>> 
>> Yes ... that should work.  Always a good idea to verify by
>> storing your own on a test...   The number of values stored
>> varies according to Torder, as I recall.
>> 
>> Paul
>> 
>> On Mon, 13 Dec 2010, nek5000-users at lists.mcs.anl.gov wrote:
>> 
>>> hi all,
>>> 
>>> i need to compute time derivatives of the velocity.
>>> Is VXLAG(:,:,:,:,1) the vx field of the previous time step?
>>> 
>>> thank you
>>> francesco
>>> 
>>> 
>>> 
>> _______________________________________________
>> Nek5000-users mailing list
>> Nek5000-users at lists.mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>
> _______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>