Two questions on GMSH

[Christophe Geuzaine]

behzad tavassoli hozori wrote:
> 
>    Dear Christophe Geuzaine,
> 
>    I have downloaded GMSH which you have
> kindly presented as public domain. It works
> very well in Win NT. Unfortunately I have
> had problems using GMSH efficiently in some
> cases. Any help in these regards would be
> deeply appreciated.
> 
>    1) Involved in analyzing passive microwave
> devices, I frequently encounter multimaterial
> domains. However I have not succeeded in
> defining such multiregions( multi-volumes)
> in GMSH. For example consider a small box
> of material A( with mesh density MDA) in a
> larger box of material B( with mesh density
> MDB). How can the user make GMSH mesh the two
> regions (i.e. A &(B-A)) so that their meshes
> would be consistent as regards to surface
> meshes on the faces of A?
> 

When defining the exterior volume, GMSH should ask you if there are
holes in it (I don't know the details of the Windows user interface,
since I programmed the Unix one, which is slightly different). If there
is a hole, you should then select the boudary of this hole. This will
define (B-A). 

The definition of A is trivial (no holes in it).

>            *******************
> 
>    2) The second point is the way GMSH allows
> the user to control the mesh density. Consider
> a rectangular waveguide( box). How can the
> user make GMSH create a graded mesh for the
> box, with nearly minimum tetrahedrons, with
> the following mesh density:
> 
>    2-1) MDA for the internal points.
> 
>    2-2) MDB for the points on two opposite
> faces of the box, so that MDB>MDA.
> 
>    This means any of the eight main vertices
> of the box will be on one of the two mentioned
> faces. For GMSH, this means the mesh density
> will be equal to MDB everywhere in and on
> ( the faces of) the box. In other words, one
> can not have the desired mesh density MDA for
> the internal points. Is there any solution to
> this case which does not require adding any
> auxiliary points( with mesh density MDA).
> 
>    This is the case when analyzing a multiport
> waveguide network. As you might know, the mesh
> density on the ports should be much more than
> that of internal points.
> 


With the version available on the web, the only possibility to specify a
mesh density without adding other control points is to define a volumic
background mesh (e.g. from an error map, or whatever).

Best regards,

-- 
Christophe Geuzaine

Tel: +32-(0)4-366.37.10    mailto:Christophe.Geuzaine at ulg.ac.be
Fax: +32-(0)4-366.29.10    http://www.montefiore.ulg.ac.be/~geuzaine/