I'm trying to get some bird feather simulations running with vellum but am getting some really odd behavior. Some of my proxy feathers are flipping upward for no apparent reason when they should be settling. There are a few spots where one feather is penetrating another, but this isn't the case throughout the majority of the region where the flipping is occurring. I've tried setting the edge-length-scale to a very small value that I believe should eliminate the possibility that the geo is being pushed up by the presence of overly high thickness, but that doesn't seem to help. There are also no forces besides gravity and wind set to zero. Can anyone here suggest an explanation and solution for this behavior?

Hard to say withou an example, but are the feathers moving/ do you have enough substeps? Also how do you fix the feathers position? A simple pin? Maybe the stopped attrib works as it can prevent rotation.

-> Vellum Attribs | state [www.sidefx.com]

@Xue_Yue , unfortunately, I can't provide an example file. I just thought it might ring a bell for someone. But, as far as collision passes and substeps, I've been using 25 and 6 respectively. That should be enough. That said, I think I've identified a major contributor to the problem. The root group that I was pinning was a row of points. When I switch the pin to instead target a single point representing the very center of the stem, the problem goes away. I can't be certain this is the only contributor to the problem as there were other changes I'd made for that same test, but I suspect it's the main if not sole one.

should be fine to pin multiple root points, but

- try to use high stiffness soft pin constraints instead of permanent or stopped

- make sure all root pinned points have external collisions disabled ( i@disableexternal = 1; )

- if you are using substeps make sure your target grometry as well as collision geometry move smoothly between frames, so they are either live deformed or properly interpolated between frames

- make sure none of the pscales is larger than half of neighbour edges, which you already said they shouldn't be

@tamte Thanks for that. External collisions are off for the root points but I don't think the simple test anim I'm using (an alembic import) has any subframes. That and the idea of using soft-constraints are good points. Thanks.

fbb
but I don't think the simple test anim I'm using (an alembic import) has any subframes

it doesn't need subsamples, but it just needs to interpolate nicely between frames, which alembic would do by default for consistent topology geo
but it's always good to click Enable Playback at Fractional frames button and scrub to see if target and collision geo is moving properly between frames right before going to sim, as it can happen that this interpolation is lost, like when using default Timeshift with $F and Integer frames on or many other cases

FYI for everyone here, I seem to have finally found a fully functional solution. That is to pin the first row of prims rather than the first row of points. When pinning just the points, using soft or permanent, I could never get the level of stiffness that I need without the feather flipping around. But when I pin the first row of prims it behaves as you would expect both in the case of permanent and soft pins.

@Tamte

- try to use high stiffness soft pin constraints instead of permanent or stopped

Can you give any insights why you'd prefer the high stiffness soft pin over stopped? Thanks.

soft pins are able to get out of collisions or dampen some competing constraints, so it usually resolves some erratic behavior I used to see around hard constraints