From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Feb 21 2007 - 17:21:10 CST

Julien,
  I've redesigned the vertex interpolation routine with a different
implementation which runs faster and should also handle your particular
dataset more gracefully. I've posted the latest test version in the
VMD Public project "alpha" directory on biocore. See the instructions here:
  http://www.ks.uiuc.edu/Research/vmd/alpha/

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Wed, Feb 21, 2007 at 01:42:05PM +0100, Avell Diroll wrote:
> John Stone wrote:
> > Julien,
> > I've reproduced your problem, and it is as I suspected, the isosurface
> > is ill-defined at an isovalue of zero. The isosurface extraction
> > code in VMD includes a safety check that prevents it from drawing triangles
> > where the interpolated vertex coordinate is calculated under conditions
> > leading to numerical instability. If I disable the safety checks
> > your surface will render without the holes, but other ill-defined datasets
> > could render badly or cause a runtime floating point exception due to
> > division by a value that's nearly zero. If you like I would be happy to
> > provide you with a binary that disables the safety check for the time being,
> > so that you can make your figures, while I see if I can implement the safety
> > check just slightly differently to allow you to get your surface even though
> > this specific case is definitely marginal at best. Try rendering the
> > isosurface for isovalues 0, 0.0001, and 0.000001 and you'll see how
> > small the volume gradient is at the outer edges of your box.
>
> Hi John,
> It's great to know there's a rational explanation to this behaviour.
> Thanks for your time spent in solving this.
>
> I would be glad to receive a VMD "safety check free" version.
>
> However if this case is marginal, it might be easier to add a console
> option to disable the safety check "manually" than to modify its
> implementation.
>
> I observed the isovalues you were suggesting and the gradient is
> definitely small on the edges (1e-17), but having a display of a
> "lipophilic transition" plane would be really nice even if the grid
> values are below the precision level of the calculation.
>
> Thanks for everything
>
> Julien Lefeuvre

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078