How resimming works?

[TomGore]

Three questions regarding resim:

1. When I resim in order to make grid higher res and to add wavelet turbulence, do dynamics/conservation/randomize settings (changing them) affect the resim result? I can't find info on that anywhere except steps should be left as 1 with base as well as resim. I sometimes get impression that my wavelet sims are going off the flow in comparison to underlying base mesh - like some values are added for the second time. Fluid tends to get thinner or thicker.
2. Does messing around with source settings while resimming with wavelet affect the resim?
3. How to add more particles when base cache is okay. What exactly should I chose or boxes to tick...

[georgi.zhekov]

Hello,

In case any settings differ between sim and resim, especially the ones under Dynamics, the resimulation will be very different. So it's best to leave everything as is when resimming.
For the particles if you wish not to do another simulation you can use the Count Multiplier in the Particle Shader - bumping the value up will create more particles at render time, though the result won't be as good as when you have the particles in the simulation itself.

[Svetlin.Nikolov]

Hey,

Some more info - everything that affects velocity will be ignored during resimulation - the velocity gets read from the cache files, so the conservation, vorticity or randomization will be ignored. However, the visible fluid quantities, such as voxel smoke amount, temperature, or drag particles amount, still need to be emitted in the simulation, so most of the source's controls will work. The steps per frame or the advection type would also play a role. You could get the visible quantities thin out if you have more dissipation in the resimulation than the original - e.g. if you have more steps per frame - each step basically blurs the voxel grid; or if you are doing a time-bend resimulation slowing the sim down - then you will have more steps again and more blur, or if you have smoke dissipation, it will do more passes than in the original sim.

Hope this helps!

[TomGore]

Thanks, it explains a lot to me...

[TomGore]

One more thing about looping mode. How exactly, from technical point of view, an "advection origin" channel and blending mode works? It seems to do a great job when looping, I am curious what goes on under the hood then...

[Svetlin.Nikolov]

Ah, the Precise tracing uses the advection origin data to try and mimic what the simulation advection algorithms do, but without having the overhead of the real simulation - this is why it only works with cache that came from Phoenix.

[TomGore]

Is it based on some actual scientific papers out of some Siggraphs methods? There are some examples of similar approaches, have you seen other methods by different authors?

[Svetlin.Nikolov]

Negative, it's our own method - we know what needs to happen to our data so that it blends smoothly and this is a big advantage over general-purpose methods that don't know the nature of the simulation that generated the cache files.

Cheers!