Computer Graphics World


Issue link:

Contents of this Issue


Page 9 of 59

8 Big Strides in Small Fluid Flow I f you've done any visual eff ects work involving fl uid, you probably look at high-speed movie clips of splashes with a tear in your eye. CG fl uid eff ects have struggled to achieve the gorgeous, rich level of detail we see constantly around us. CG has been revolutionary in achieving realistic fl uid motions, optical properties, and directability. But for years, we've hit a wall when it comes to super-realistic levels of detail. Studios tend to face this issue at the large-scale end of the fl uid VFX spectrum, for eff ects like submarine breaches and fl oods in New York City. Driven by big-budget fi lms, large-scale fl uid VFX moved forward signifi cantly in the past decade, through studios doing smart software development and, of course, through more industrial-strength hardware, basically all to jam in more particles or more voxels per pixel of screen space. Where fl uid VFX has gone almost nowhere is in the opposite direction of scale. But that's just where we've now made big strides. We're talking about tiny fl uid VFX, like "crown" splashes from raindrops falling into puddles or splats of paint hitting a wall. Ironically, these are the most commonly needed types of fl uid vi- sual eff ects, and they're what CG has been worst at achieving. Look at real versions closely, and you see features that are constantly requested for commercial projects—and invariably end up coming from table- top shoots because CG fl uids don't make the visual bar. Worst of all, these elements are always meant to be heroic, gorgeous, and graceful. Usually this means slow motion. No cover-ups with motion blur. No- where to hide. Let the migraines begin. e challenge of tiny fl uid eff ects has two causes. First, there's the same problem as for large-scale fl uid VFX: resolution. A graceful sheet of water arcing through the air may be small in size, but it contains billions of "particles" (molecules) that grant it huge amounts of detail. When the camera is close and the frame rate is high, you see that detail, at which point the splash might as well be from a breaching submarine. Second, the physics diff er from large-scale fl uid VFX. When we change the scale of observation by orders of magnitude, at the big scale, forces that were insignifi cant now can drive the fl uid motion. For fl uids, small-scale forc- es that start to become important are intermolecular—for example, elec- trostatic and chemical—some of which get called simply surface tension, and these are at best only poorly simulated in CG fl uid engines. e issue becomes clear when you try to get CG fl uid to do some- thing simple: fl y through air as a delicate, thin, continuous sheet, for instance—you know, like in all those reference movies. Here's where we generally hit the resolution issue. If using a conventional SPH (smoothed particle hydrodynamics) fl uid solver, such as Next Limit Technology's RealFlow, those thin sheets just aren't stable. In a thin sheet of SPH fl uid, say one or two particles thick, you only need the particles to jostle slightly away from each other and you quickly develop holes. e solver has no capability to allow the sheet to thin out because you're at the resolution limit—unlike in real water, where those sheets are many molecules thick, and on top of that, the micro-forces that come into play (such as electrostatics) hold the sheet together for at least a little while. e fl uid simulator has trouble approaching the needed fl uid resolution, and it doesn't include those needed special forces. e result is you get fl uid with a lot of Swiss cheese-like holes that is decid- edly unattractive to every creative director I've met. Luckily, workstations and fl uid simulators like RealFlow have got- ten much faster, and many tools have also become customizable as the developers open up more of their calculation engines to users. You can write your own plug-ins to add behaviors to the base fl uid simulator, which we do constantly at Fusion CI Studios to get custom behaviors. Because we do a lot of commercial work, we often face the challenges of small-scale fl uid VFX, so we've recently developed an algorithm that prevents SPH fl uids from developing into Swiss cheese. We refer to our in-house tool as Smorganic (smooth organic) be- cause it allows SPH particle sims to be smooth (no holes), but does so dynamically during the simulation, permitting the fl ow to continue in a natural way. Other techniques we developed earlier involved unnatural tactics, such as creating dominating forces to hold the fl uid together, or post-simulation processing to reduce the size of holes. ose meth- Simulation By MARK STASIUK and LAUREN MILLAR Lauren Millar and Mark Stasiuk are co-founders of Fusion CI Studios (www.fusi-, which specializes exclusively in photoreal CG fluid and particle effects. In recent years, facilities have worked on large-scale fl uid simulation, and have only just started focusing on fl uid fl ow at the small scale. December 2009

Articles in this issue

Links on this page

Archives of this issue

view archives of Computer Graphics World - DECEMBER 09