>Hi there, > >Sorry if this is a newbie question, but... > >I'm trying to visualize a 3D grid (of charge density) >which is not rectangular, but does have regularly spaced data >points. (It's actually a hexagonal grid.) In particular, >I'm having problems getting the data prompter to accept my >deltas (which need to be vectors rather than scalars) in >its input fields. > >For example, I'd like to set-up a grid having the >following characteristics: > >origin = 0,0,0 >delta_1 = 1, -1, 0 >delta_2 = 0, 1, 0 >delta_3 = 0, 0, 1 > > >Is such a thing possible? >
The idea is interesting but unfortunately doomed by lack of a "hexagon" connections element in DX (currently: anyone want to write one?). Also unfortunately, there is no compact representation of such a mesh of positions. The origin/delta/counts scheme can construct regular orthogonal or skewed position meshes only. After that, you must write out the list explicitly even though we perceive a regular structure in a hex tiling. The best you can do with what DX has is delta_1 1,0 delta_2 0.7,0.7 For 2D, this will create vertices that could be thought of as the border and center of a hex pattern, but of course, it only covers a domain in about half of the positive X, positive Y region. (by the way, if you are describing a hexagon, I assume you mean 2D, therefore you should not designate a 3rd delta which tells DX the object is 3D; if you are describing 3D, I think we need to talk about hexahedral objects (also not currently supported)). The only connection elements available natively are: (no connections) for scattered data lines for 1D triangles and quads for 2D cubes and tetrahedra for 3D theoretically, cubesNd for >3D (never saw one myself, though if myself yesterday got together with myself today and tomorrow, maybe we ARE seeing one!) also, the less useful Faces,Loops,Edges (better for GIS type applications) So, you will need to create triangular subdivisions of your hexagons, which of course means writing out a long series of integer 3-vectors indicating the position indices (0-based) of the vertices of each triangle. You only need list each position in the position list once, but on average, you will use the same position index about 6 times (depending on how you triangulate) (kind of hard to draw an example, but if you draw a hexagon, pick a vertex and connect it to all currently non-connected vertices, you subdivide it into 4 triangles. After tiling the surface, some vertices are shared by 8, some by 4, and of course edges are 1,2,4, or 6). If your data is position-dependent, you don't need to do anything else to visualize it. If connection (hex) dependent in your data set, you'll also need to assign the same data value to the 4 internal triangle elements, thus expanding the size of your data array by 4x. This is all done by a custom filter program that you write that outputs a DX native file format containing positions, triangular connection elements, and data that maps either to position or triangle. You can't do this degree of restructuring within DX (if you can, I don't want to think about how difficult it would be). The above scheme has the advantage of reducing the number of elements but introduces a visual bias since each triangle only knows about the color (data) of its 3 vertices. So if the data values vary around a hexagon, you'll see discontinuities along the edges of its internal triangles when colormapped. You might prefer to create a center point for each hexagon, connecting 6 triangles from center to rim vertices. Then, you would see less colormap artifacting around the center point (I think) at the expense of having to create 50% more elements but only 25% as many new data values (an average of the hex points assigned to the center point) than in my previous scheme. I know it sounds painful, but if you want to use the many visualization capabilities that depend on interpolation (like colormapping and isosurface), you must provide a mesh in a native DX format. Finally, though I'm not well versed in it, there is a native "neighbors" component that might be useful to help you traverse the adjacent hexagons, but again, I believe you will have to construct this component yourself and probably custom modules to do anything useful with it once you have it. Maybe others can suggest uses for it using native modules. Chris Pelkie Vice President/Scientific Visualization Producer Conceptual Reality Presentations, Inc. 30 West Meadow Drive Ithaca, NY 14850 [EMAIL PROTECTED] (607) 257-8335 or (607) 254-8794
