On Aug 26, 2013, at 10:17 AM, Daniel Wheeler <[email protected]> wrote:

> I've also had issues with small meshes in Gmsh. This is obviously a 
> completely dumb problem. The absolute length scale of the mesh should have no 
> bearing on Gmsh's outcome. Anyway, in the past I've just scaled the problem 
> up so that the grid spacing is ~1. You can then rescale the mesh in FiPy or 
> rescale the equations.

This can be an issue, but I'm not clear that it is here. The problems I usually 
look at have features on the order of nanometers to micrometers. 2 cm x 1 cm 
sounds astronomical to me. One question is what units are you using? In CGS, 
then your mesh is order 1 as advocated by Daniel. Even in MKS, I can specify 
sub-nanometer grid spacings without much difficulty.


On Aug 24, 2013, at 10:07 AM, John Assael <[email protected]> wrote:

> Could you please kindly suggest me any settings that could balance precision 
> and performance for such a small size?

You can solve at some uniform resolution and then if you can calculate an error 
metric for that solution (there are a variety of ways and they depend on what 
you're calculating), then you can use that error metric to scale the mesh by 
passing a "background=" argument. See the end of the docstring at 
http://www.ctcms.nist.gov/fipy/fipy/generated/fipy.meshes.html#fipy.meshes.gmshMesh.Gmsh2D





_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to