[Gmsh] edge length * 0.5 => number of elements * 6?

[Martin Vymazal]

Hi, 

 I tried to use 'Refine by Splitting' (instead of changing the clscale) on two 
different meshes, and the factor I got was exactly 8.  There is also at least 
one way of splitting a tetrahedral element into 7 (not 8) smaller tetrahedra - 
the subdivision is not unique.

Best regards,

   Martin


On Tuesday 02 July 2013 17:02:15 Geordie McBain wrote:
> 2013/7/1 Nico Schlömer <nico.schloemer at gmail.com>:
> > for a tetrahedral mesh, when scaling the characterstic length by 0.5, I
> > notice that the number of tetrahedra roughly increases by a factor of 6.
> > Is this generally true? Does anyone know a reference for this?
> 
> I would have guessed à priori that the factor was eight (one over the
> cube of the length-scaling).  My numerical experiments with the
> attached simple cubic geometry aren't conclusive, but the number seems
> a bit short of that, though higher than 6.
> 
> $ gmsh -3 cube.geo -clscale $CLSCALE | grep elements
> 
> CLSCALE elements ratio
> 1.0              2 289        -
> 0.5            15 605    6.8
> 0.25         114 034   7.3
> 0.125       748 859   6.6