RobynsVeil opened this issue on Dec 03, 2010 · 409 posts
aRtBee posted Mon, 21 May 2012 at 6:13 AM
@EnglishBob
as far as the belt-thing goes:
the collision offset should be less than half the (least) distance between the belt and the body
if the belt is two-sided from a mesh perspective (so it had vertices, edges, polys at the inner as well as the outer side) do try to reduce the collision depth to (preferably half) the thickness of the belt - then the cloth cannot 'see'the outside. A trick can be to split the belt into an inner and an out group, and make the cloth collide to the inner group only.
follow the reipe I just posted. I guess you will need quite a large Steps per frame setting (in a recent test i needed about 3000 to prevent a lowres cloth to fall through a hires box with a thickness of 0,005% = 0,013mm).
to my understanding there is no algorithmic preference between in and out. But... belt, cloth and body might have very different mesh densities. Most cloth items I see passing by have a vertex distance of say 2 to 5 cm (1-2 inch), while Vicky has an average vertex distance of say 0.5cm (1/6th inch) only. Cloth agains body means: medium / lowres mesh against very hires mesh. Belt against dress means: lowres against medium. People tend to forget the effects and differences of that.
as far as Delauny goes: the math theory on advanced cloth sims state that the most regular meshes are the least fit for purpose in this area, so that's promising for Delauny 9and bad news for quads). For further reading into 3D cloth sim, I can recommend http://www.artbeeweb.nl/researchpaper_manasilib.pdf (10Mb).
have fun.
- - - - -
Usually I'm wrong. But to be effective and efficient, I don't need to be correct or accurate.
visit www.aRtBeeWeb.nl (works) or Missing Manuals (tutorials & reviews) - both need an update though