I’m working on a fairly complex smoke simulation to create a sand storm in 2.8. I’m almost done, but one of the things I noticed about quick smoke simulations in general is that smoke is emitted in waves. I was trying to figure out what was doing this in my simulation and I also noticed it in the even the simplest quick smoke setups. The emitter gives off pulsed waves of smoke rather than smooth and continuous output.
Does anyone know what creates this?
Do you mean the pulse at the start of the sim? Not sure, but my guess it that it comes from the strong temperature gradient at the start. The temp of the air is set to default, while the emitter is hotter. Until there is more of an equilibrium, strong acceleration away from the emitter will happen in its vicinity.
Or do you mean the curls appearing at the rim of the upward flow? These come from the vorticity confinement used in this old sim method. Set vorticity=0 and it should be gone. More modern methods like FLIP (as used by Mantaflow) don’t do it.
I mean the periodic expansion waves appearing in the upward flow. If you setup a simple quick smoke and start the sim you can see pulsing phenomena. My observation is that the frequency is about once per second (from memory).
I’ll experiment with the vorticity setting and see how that affects it. But what if I want vorticity?
I’ll also look into using Mantaflow. I was watching some of their simulations and there is a definite improvement in quality.
Thanks!
Okay. I see.
The weird pulsing has nothing to do with the emitter, but appears by itself at the rim of the upward flow. It is an effect of turbulence appearing out of a laminar flow (but an unphysical one here, as explained below).
With that question you are exactly on point. It shows what exactly is the problem with the used algorithm. These (simple) grid-based methods add extra (mostly undesired) viscosity, which damps away vorticity/turbulence. Thus there is used a trick (named vorticity confinement), which adds again some (unphysical) vorticity. This effect is what the vorticity parameter controls.
Such an upward flow is a situation where it becomes very apparent that the added vorticity is unphysical, looking really unrealistic. But it is the only option the sim has.
Mantaflow uses a particle-based algorithm which doesn’t add the extra viscosity, so there is also no need to add extra vorticity. Thus the sim works more physically correct, and you can see that.
Thanks, omgold. This is a situation which has plagued my sim because I’m simulating blowing sand and the weird flow artifacts forces me to do all manner of strange things to get a uniform blowing sand at specific scales. I either have to disable vorticity which makes the sand storm look unrealistic close up or increase my focal length on renders which looses much of the turbulence the storm should have.
I could play tricks all day, but I think it might be better with mantaflow. It looks easy enough to translate the simulation to mantaflow. Again thank you.