[Nek5000-users] prex mesh precision

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

Hi Frank,

The main objective of going high order is to minimize dispersion
errors so that the error at the end of the simulation is at
acceptable engineering accuracy.  Dispersion errors accumulate
linearly with time and thus, for long time-integrations, one
requires that the leading contribution to the (discretization)
dispersion error be small.  Roughly speaking, for a final
error of .01 in and a final integration time T=1000, one would
want the error to be 1.e-5.   Such tight tolerances are fairly
easy to realize with a high-order discretization.

That being said, there are times (e.g., stability computations)
where you really want to control all the errors in order to 
establish critical Re or Ra to several digits.   In such cases,
I typically project the geometry from prex onto the true geometry.
This is easy for spheres and cylinders and only modestly more
difficult for other shapes.   I can help you with some code to
do this if you like - it involves inserting lines of code in either
usrdat() or usrdat2()  [ which in fact is why those stubs were
originally created ].

Cheers,

Paul


On Sat, 10 Jul 2010, nek5000-users at lists.mcs.anl.gov wrote:

> Hello all,

I was wondering what would be involved in getting "prex"/"prenek" to
perform all calculations in double precision and to output a mesh in
this precision.  In the cases we are looking at which involve
understanding stability boundaries and growth rates, the high accuracy
of NEK is the prime reason for using it.  There is some concern that an
inaccurate representation of the boundary could potentially be a source
of meaningful error, in particular in the course of grid refinement
studies.  Putting my two cents in, it seems somewhat incongruous for the
fundamental geometry definition for such a highly accurate method/code
to be in single precision.

Cheers,
Frank


-- 
Frank Herbert Muldoon, Ph.D. Mechanical Engineering
Technische Universität Wien (Vienna Technical University)
Inst. f. Strömungsmechanik und Wärmeübertragung (Institute of Fluid
Mechanics and Heat Transfer)
Resselgasse 3
1040 Wien
Tel: +4315880132232
Fax: +4315880132299 
Cell:+436765203470
fmuldoo (skype)
http://tetra.fluid.tuwien.ac.at/fmuldoo/public_html/webpage/frank-muldoon.html

_______________________________________________
Nek5000-users mailing list
Nek5000-users at lists.mcs.anl.gov
https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users