Question regarding slow down behavior - strange movement

[olika]

Hi Svetlin,
at first i have to point out how great your support is! bigup!

I was already fearing we`re coming to that point, where interpolation has its limits ;-(

So for me it totally makes sense what you`re saying. Sad but true hehe
Few new thoughts:
1. see new gif, it`s simulated with the slowdown method again but this time i unlinked the grid from the bike (even though it was linked within the base sim)
Awesome how smooth this result is compared to the general-purpose method
2.strange behavior somehow related to the linking?
3. I had a slo-mo it the last project with no problems, in that case it was a car hitting little puddle. So it was possible simulating the whole hitting process in slow mo without any simulation upfront.
therefore no interpolation, nice slo.mo, all good
But this time it`s needed that the bike creates that big pile of dust first, and then somewhere in the middle we want go into that slo-mo.
So i`m scratching my head what would be a intelligent approach to that scenario?
A simulation of the whole 150 frames with 25 fps would mean 1440 frames simulation with 240 fps slo-mo and that with a 100 mio grid uh uh that hurts...

[Svetlin.Nikolov]

I do suspect the linking to be the reason for the difference. Still haven't proved it yet though. Also, the loss of detail is expected with the Slow-Down method - it runs an actual simulation and the higher the slowdown factor, the more steps per frame the simulation performs, and thus the more numerical dissipation, leading to blurred smoke.

Will be back with more updates...

[olika]

Crossing fingers!

[Svetlin.Nikolov]

Hey, so I think it's a combination of crossing the limitations of the algorithm and getting the speed and magnitude of all interacting forces and particles reduced proportionally. Here are the changes we came up with for slowing it down 10 times:

- change play speed to 0.1
- change animation speed of the moving bike - place a new key at frame 0. There we have a value 34.1 and goes up to 100, so its length is 66, which must become 6.6. Thus, at the final frame we should have 34.1 + 6.6 = 40.7.
- wind turbulence 30->3
- pflow emit stop 180->1800
- pflow both forces influence 1000->100
- pflow speed 8.0->0.8; variation 2.0->0.2
- pflow delete by age 9+/-3 -> 90 +/- 30
- pflow delete by age 9+/-1 -> 90 +/- 10

However, it won't look good with such slowdown anyway. We are researching more robust algorithms for retiming simulations, and I hope in a couple of months they will be ready to use in the nightlies.

Cheers!

[olika]

Hi Svetlin,
thank you so much for all the support, that`s just incredible!
Luckily the whole thing is a private ting and not a client wish hehe!
So hey just skip that slo-mo thing in that particular scene and i will try again with what you will come up in the future in a new scene.
I like the results even without the slo-mo (that would have been just the topping ;- ) ) and maybe it`s just an aweful scene to try that.
Here you can see the results i`ve been fiddling on, just posted it in the image uploads:

[Svetlin.Nikolov]

Sorry about the bad news - I hope we can have a better method to do that soon. The setup looks cool