Smoke amount emitted tied to cell number?

[ben_hamburg]

Hi there,

I have another issue i can´t quite figure out: I´m currently running a smoke sim, where smoke is constantly beeing emittet inside a tube.
I first did a relatively low-res sim (500k cells or so) and once I was happy with the dynamics, I ran another sim with around 30 mill cells.
But now the sim looks different, it seems like there is less smoke beeing emitted overall and the smoke isn´t crawling through the tube as fast and as far as it did before.
How is this happening? Does it have to do with the conservation method (direct symmetry), the material advection (Multi-Pass) or do I have to increase the step size when I increase the cell number? Or was the look of the preview just deceiving, because of the bigger triangles?

[ben_hamburg]

I´ve compared the simulations now and maybe its an issue with my collision geometry...there are small crevices in it and when I decrease the grids resolution they might not be considered by the simulation, when I increase the resolution it looks a bit like the smoke gets lost in those crevices...

If that is the case: Can I prevent this somehow? Or can I work around it by doing a lores sim to get the dynamics right and then doing a resimulation to add detail? I´ve read about this topic in the manual but I´m not sure on which options to use in this case, so please check what i would do:

1. Simulate lores sim and don´t forget to check the "wavelet" channel (do I also need to export "velocity" for this?).
2. Check "Enable" in "Grid resimulation".
3. Amplify resolution (If I would have increased my grid resolution twice, I would instead set this value to "1")
4. Choose "Wavelet fast" or "Wavelet nice".
5. Leave other values at their default.
6. Use original simulations path for input and set another path for resimulation output.

[Svetlin.Nikolov]

Hey,

This is the proper workflow, yes - you do need the velocity from the low-res sim in order to be able to resimulate. The thing with increasing resolution is that by changing the voxel size, the dynamics of the fluid will unfold in a completely different way. Even by changing a single voxel to an obstacle would produce a butterfly effect and the sim will be nothing like the previous 50 frames later. This is why resimulation is the way to increase the resolution while keeping the overall shape and dynamics - it will force the high res fluid to follow the low-res velocity from the original sim.

Sorry for replying so late. Do you have progress on this?

Cheers!

[ben_hamburg]

No problem, I´ll try that the next time. For this time the problem kind of fixed itself and I´m not sure what I did, I think I increased the stepsize and then the smoke didn´t dissipate as quickly. Or it was just the opacity-curves again...