Hi all 🙂
Stability is one of the most desired feature of a simulation and in particle systems is a must have. The fluid particles exhibit random explosions (which I think could be reduced because in older SVN revisions that does not happen) and in general the particle systems have a leaking trough wall containers behavior when the container rotates. (think of the typical wine glass falling example 🙂 )
As a suggestion from Jahka and several members of the community I have implemented particle subframe simulations not just for particle fluids but for the whole particle system.
Subframe simulations greatly increase the stability of the simulation to any desired level, it basically perform a user defined number of pre-simulations in between frames, allowing finer grain sims, and a nice side effect of it is that now particle-surface collisions could be properly captured so there’s no leaking effect when containers rotates.
Subframe simulations do increase the calculation time proportionally to the number of sub-steps, so the user has to made a trade-off between stability and simulation time. But the good news is that there’s now a lot of methods to reach stability in the particle systems so a combination of them will help to get very low substeps (even none) to get stable simulations and very fast proper particle collision detection.
Here are some examples:
2 subframes (you don’t even want to see no subframes 😉 )
Fluid in rotating container, with no subframe
Fluid in rotating container, with 5 subframes
Newtonian particles with a rotating container and no subframe
Newtonian particles with a rotating container and 4 subframes