Properties to change when scaling down a soft body

[josemendez]

Hey,
So I've been experimenting with a soft body tongue around the same size as the rubber dragons. I've gotten it to feel how I want it to, however I'm having a difficult time reducing it's size and keeping a similar behavior. I'd like to scale it to around 1/20th of it's current size and would like to know the parameters I should change (and possibly what to change them to) in order to maintain a given behavior. My existing tongue at a scale of 300x/300y/300z are:

particle radius: 0.1
overlap: 0.2
smoothing: 0.5
anisotropy neighborhood: 0.2
max anisotropy: 3
soft cluster radius: 0.3

I've tried reducing the tongue to about 1/20th of the above scale and modified the particle radius, overlap and smoothing, but it seems like the tongue still acts very stiff. What are some other properties I can change and any suggested values for reducing it to 1//20th of its size? Thanks!!

Hi,

It's very difficult to ensure consistent behavior when changing the size of an object in world space, since Obi is not a continuum-mechanics based solver, but relies on lumped-mass approach since it's simpler and faster to simulate. So changing the particle/cluster distribution affects material stiffness.

You could use local-space simulation and scale the entire solver instead (though this comes with a bit of a performance hit):
http://obi.virtualmethodstudio.com/tutor...space.html

If you wish to stay in world-space: in any iterative solver low stiffness can either be intentional, or a byproduct of slow convergence. This means that you can either spend a lot of time (iterations) in your simulation to ensure the exact stiffness you set is reached, or you can just bail out early each frame and use the "error" (that manifests itself as under-estimated stiffness) to your advantage .

These are the factors that affect stiffness:

http://obi.virtualmethodstudio.com/tutor...setup.html
- Smaller particle radius -> more particles-> less stiffness (since there are more equations to work with, the simulation converges slowly)
- Smaller cluster radius -> less stiffness (less overlap between clusters, so stiffness propagates less far away from each particle)

http://obi.virtualmethodstudio.com/tutor...olver.html
- Less shape matching iterations -> less stiffness (again, because less iterations per frame reduce convergence speed)
- Using parallel evaluation instead of sequential evaluation -> less stiffness (parallel mode averages results from each shape matching cluster, "spreading out" stiffness)

http://obi.virtualmethodstudio.com/tutor...aints.html
- Less stiffness -> less stiffness (this one reduces the actual material stiffness, all others decrease stiffness because of slow convergence)

This can sound quite technical if you're not experienced with physics engines, so drop me a line if you need help.