Hi Ned,

Thanks so much for the input.  Chris P. and I have been discussing
this off-list and he's provided a number of ideas (such as
CollectMultiGrid, etc.), but we haven't yet resolved the issue.

On Fri, Mar 14, 2003, Ned Piburn wrote:
> Daniel,
> 
> I'm puzzled why you don't have boundary nodes on your, 
> well - er - ah, on your boundary.

Hmm. I'm sorry, but I don'm not sure what you mean by this.  The slice
the you see at http://www.physics.cornell.edu/~freedman/OpenDX/pbc.png
shows the mesh that runs up to the boundary at the PBC edge.  I think
Chris was surprised that I had a gap between the values, but as far as
I understand it, I don't have a gap.  That gap is the same spacing as
the normal regular grid of my data points (position- not
connection-dependent), it's just that the connections are not present,
as I'm sure you're aware.

> I do a fair amount of this sort of thing, in Euler and 
> Lagrange frames of reference and cylindrical and 
> Cartesian geometries.  For the simple Cartesian case, I 
> use the SCALE module with a -1 in the scaling vector to 
> reflect across the appropriate plane.  As you point 

Is this any different than using a translate or a mark-compute-unmark
series?  Or is it just another method to keep in mind to do a similar
thing?

> out, this does not produce a new contiguous mesh.  It 
> must be collected with its progenitor and rendered. 

Okay, but _this_ is the problem area for me.  I know how to collect it
with it's non-translated ancestor, but I can't get the isosurface to
span the gap where the connections are not present.  Do you have any
ideas how to do so?

> There are couple of other gotchas with this technique: 
> 1) Normals generally to not get recalculated as they 
> ought to.  So what I tend to do is delete the normals 
> (REMOVE module) and then recalculate them (NORMALS 
> module).  2)  If you have a surface on your plane of 
> symmetry and then you reflect across that plane, the 
> outside surface will have a seam in it.  Remove that 
> extra surface on the plane of symmetry and the seam 
> will go away.  Note that ISOSURFACE does not generate 
> mesh boundary surfaces, you generally have to go to 
> some extra trouble (SHOWBOUNDARY, BAND, INCLUDE) to 
> create them.

Okay, I don't think I have to worry about this problem, since I don't
use normals and don't have surfaces on my plane of symmetry.

Anyway, if you see in my vis pic:
http://www.physics.cornell.edu/~freedman/OpenDX/pbc.png

Essentially, I want to find a way to connect the half-spheres.
Any help you or others might provide is very much appreciated.

Best wishes,
Daniel


> -Ned Piburn
> 
> 
> Daniel Freedman wrote:
> > 
> > Please excuse my resending this.  I'm not sure the first email got
> > through to the list.  Thanks for your help! -Daniel
> > 
> > ------------
> > 
> > Hi,
> > 
> > I've search the OpenDX reference, quick start, and mailing lists, as
> > well as read VIS's "Paths to Visualizations" (which is quite helpful),
> > but cannot find any guidance on this matter.
> > 
> > I want to translate (or copy--which is just a tranlate then collect
> > original plus translated) a data field, with regular cubic positions
> > and connections, by the size of my bounding box to better visualize
> > my system which physically has periodic boundary conditions.
> > 
> > I can think of two easy ways to do this:
> > 
> > 1. Mark ("positions"), Compute, Unmark ("positions")
> > 
> > 2. Translate module
> > 
> > The big problem with both of these (which I had hoped the Translate
> > module would fix, even if Mark-Compute did not, due to Translate's
> > more specific function), is that the connections are broken across the
> > border between the old and the new data set (at the periodic boundary
> > condition).  This is easily seen with the ShowConnections module.
> > (For the next few days, I'll leave a demo of this exactly at:
> > http://www.physics.cornell.edu/~freedman/OpenDX/pbc.png)
> > 
> > The greater difficulty with this, of course, is that a subsequent
> > isosurface across the periodic boundary condition shows a noticeable
> > and detracting void, where the missing connections should exist.
> > 
> > I'm actually quite surprised not to have come across additional
> > references to this scenario, as (at least from my vantage point) it
> > appears quite common and the solution seems non-obvious to me.
> > 
> > I'd appreciate if folks have guidance either on how to properly
> > translate to handle the connections across the PBC, or to also
> > Mark-Compute on the connections field to similarly translate them
> > (it's not clear to me how to operate on the mark'd connections product
> > mesh to get it to Do the Right Thing).  It seems to me that explicit
> > creation of the connections field, via a Regrid module, would be quite
> > cumbersome and inelegant.
> > 
> > Thank you kindly for any guidance,
> > Daniel
> > 
> > --
> > Daniel A. Freedman <[EMAIL PROTECTED]>, Graduate Fellow
> >   Electronic Structure Calculations, LASSP, Cornell University

-- 
Daniel A. Freedman <[EMAIL PROTECTED]>, Graduate Fellow
  Electronic Structure Calculations, LASSP, Cornell University

Reply via email to