@Happenstance Heroes: Thanks for the additional info! Most of my implicit models are results of cool accidents. I just keep plugging in weird combos until something cool pops up. I go through a ridiculous amount of trial expressions before I get anything usable.
There are a couple of points I think I can add to the steps you mentioned. When I first start experimenting with different expressions using the Implicit Filter in MeshLab, I use a larger voxel size (.1 or .05) until, after many attempts, I finally get some halfway decent results. Then I'll change the voxel size in the implicit filter to a much smaller number, like .01 or .02. This sometimes makes for an extremely large number of faces; sometimes over 4 million, way over Shapeways 1 million limit.
I then usually apply the
HC Laplacian Smooth filter a few times. And I'll reduce the number of faces using the
Quadric Edge Collapse Decimation filter (in the remeshing filters menu.) Then, like
Happenstance Heroes mentioned, I'll add thickness to the implied surface using the
Uniform Mesh Resampling filter. In that filter, my settings (as a starting point) are almost always set at around .35% for precision, and 50.5 to 51 for the offset percentage. And that number depends on how large the final model will be printed and how thick I need the walls. I also click on (select)
clean vertices,
multisample and
absolute distance (but not
discretize).
Under the view menu, I'll click on
Show Layer Dialog. That dialog box will show that I have my original surface (layer) as well as the new layer (the thicker model) created with the Uniform Mesh Resampling filter. I'll delete the original layer before saving it as an STL file. Then I usually open that file in NetFabb to size it correctly and to check the final wall thickness.