anamous Posted October 27, 2007 Share Posted October 27, 2007 say you have a convincing watery particle fluid simulation cooked up - anyone got any tips on getting nice looking meshes generated? the obvious method is to copy metaballs/convert/peak, but I'm kind of wondering if there is something better since the results look very "blobby" and "CG-ish" for lack of better words. Apparently, Zhu-Bridson metaball kernels would give me a nicer result, but sadly, they're not implemented. Also, I'm hazarding a guess that Osher/Fedkiw hybrid level set/particle methods would need to be implemented outside Houdini for performance reasons. SDFs? also, part of the problem is of course the fact that H9's fluids are single-threaded, so cooking sims takes ages, whereas RealFlow's multithreaded engine is quicker. This forces me to keep the particle count down (using the separation parameter on the particle fluid object) to be able to get results in any kind of foreseeable future - currently, at 20,000 particles, simulation times are 10 minutes/frame. using 16% CPU on an octocore machine. 20,000 particle is far away from any kind of high-end result. Any tips/ideas? thanks & cheers Quote Link to comment Share on other sites More sharing options...
SpencerL Posted October 27, 2007 Share Posted October 27, 2007 Check out the Particle Fluid Surface SOP Quote Link to comment Share on other sites More sharing options...
anamous Posted October 27, 2007 Author Share Posted October 27, 2007 Check out the Particle Fluid Surface SOP thanks for the quick reply. I checked it out but it seems like a wrapper for the usual copy metaball method.. still very blobby, no crisp edges, doesn't take the particle's speed/"orientation" into consideration. or maybe a matter of setting the parameters right? Quote Link to comment Share on other sites More sharing options...
anamous Posted November 1, 2007 Author Share Posted November 1, 2007 As of the newest H9 builds there's a new particle fluid surface SOP that is much better.. Been playing around with it and it demands respect coupling it locally with the old method should yield very nice results. Also, I found out the hard way that H9 on 64bit linux is a completely different program, it's greatness++. I see much more multithreading going on and of course the memory thing is just sweet. But I still need to understand this: when I simulate in relative low res for interactivity reasons, what are the parameters that I can use to get approximately the same results (of course they won't be exactly the same) in a higher resolution? Right now I control the resolution using the particle separation parameter, but the simulations are completely different (particles exploding all around the place instead of sticking together for example). Is there anything I need to adjust in unison with the separation parameter? And what about the mysterius gas constant? any guidelines for it besides the short notice in the help card? Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.