Jump to content

Pyro upres advise.


sebkaine

Recommended Posts

Thanks again for your help Paul ! :)

 

But after watching the test , i am not that convinced by the technic.

Of course when you have Jerry Tessendorf at Home that cook this with love for you , i'm sure it's impressive ...

 

But the test post , has this fume fx aspect , kinda too swirly that feel unatural to me , sur it looks nice , but like marty say going higher in res should give you better outpout.

I'm quite sure you can get stellar outpout if you really understand pyro. But like always with fluid solver , it takes lot of time and trials ...

 

I already have a lot to do with all those damn gaz solver ... :)

Edited by sebkaine
Link to comment
Share on other sites

Yes, just pushing the solver resolution often seems to be the best bet. What Jerry says in his post about using advection in a fraction of a step, rather than across multiple steps, to give the sim an extra bit of smoothness makes a lot of sense. Use it as a bit of extra seasoning, not as your main ingredient.

Link to comment
Share on other sites

Exactly Julien ! you are very close but still not enough detail.

 

a - Did you push your container res ?

or 

b - Do you use a post - process op like gridless advection ?

 

in case b i would be happy to see a little playblast , cause those post process tricks often introduce some weird stuff in the sims ?

 

Cheers 

 

E

Link to comment
Share on other sites

It's pretty much just a standard upres. In the upres sim, I multiplied the density field's divsize by 0.4 so there is even less detail loss from advection. This is quick to sim, so I could push the res a lot more.

I didn't take any time to add diffusion since it's very tricky, but I boosted the low density values in post, that helps keep the very thin density.

lowres divsize: 0.1

upres divsize: 0.025

upres density field divsize : 0.01

 

 

 

bunker_upres.mov

  • Like 1
Link to comment
Share on other sites

I pushed the res a bit more: lowres: 0.07 , upres: 0.0175, upres density: 0.007 and it looks sharper when rendered.

sim 10mins for 60 frames on 40 cores HP Z840, max ram 54GB (in interactive mode, not batch)
Upres is definately a good option I think, the lowres sim runs almost real time so you can get quick feedback.

 

post-3635-0-47319700-1455790323_thumb.pn

Link to comment
Share on other sites

That looks really better Julien !

 

would it be possible to share a simple hip that describe how you do your up-rez ?

To be sure i get things correctly when you are saying "upres" you mean by that, the process to add detail to a low-res sim with a post treatment noise advected on this low res-sim,

like the wavelet - turbulence process ?

 

So basically this ?

http://www.sidefx.com/docs/houdini15.0/nodes/dop/gasupres

 

Thanks again for your help !

 

ps : i would be curious to see a blast of the last sims to see how it handle collision. cause you're img is very close to my goal ,

i'm just afraid that the fluid doesn't look to slide on the pyramide.

Edited by sebkaine
Link to comment
Share on other sites

gasupres DOP yes, you can even get it from the shelf tools.

I use curl as a mask for noise, the default is constant for some reason but that won't get you anywhere...
You have to make sure the "curl" values are re-mapped (ramp) on the edges of the smoke, just turn on the visualization tab.

The collisions are only as good as your lowres sim so if the smoke doesn't slide perfectly, you can control the velocity near the collision boundaries.

you get a nice SDF field for free: f@collision, the values are inverted though, negative values outside.

I used it to add extra disturbance near the collision. Also there is no need for temperature.

I'll try to render and create a HIP if I have time tomorrow. 
 

Edited by bunker
  • Like 2
Link to comment
Share on other sites

Hi, thanks for great example.

How would you simply describe how upresing works? 

 

The only thing that is needed from low res sim is vel field and source.

Is it basically advecting higher resoultion source density with low res vel field? And adding turbulence to upsampled vel field to introduce smaller detail. (with high frequency so that overall shape is the same).

Does this step even need non divergent projection? (vel field from low res sim should be divergence free, but new vel with turbulence may not be divergence free)

 

So if I understood it right then it is the most efficient way of doing simulations? Because the most computation heavy tasks are non divergence projection and advection. And if divergence projection is computed on low res sim then it is quite optimized workflow. 

 

Juraj

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

@Juraj you’re correct, the lowres velocity is up-sampled, curl noise is added to it so it’s still divergence free,
density is sourced into a hires field and advected using the hires vel - the projection is done in the lowres sim.
It’s a much more efficient way of working but it requires a good looking lowers sim first of course.
You’ll also need a lowers density field for the upses to work if you’re using density to remap the turbulence.

I would recommend creating you’re own upres solver or at least tweaking the default one, that way you’re not limited
to one frequency of curl noise and you can add disturbance (random noise per voxel) and get even finer details,
it's not divergence free but the amount is negligible.

 

 

 

Edited by bunker
  • Like 5
  • Thanks 1
Link to comment
Share on other sites

  • 4 weeks later...
  • 2 years later...

I really have to bump this thread - I found this particular example of the up-res-setup to be one I will be using in any pyro I do from now on I think. I think it's amazing!

Thanks so much Julien!

Also played around with comparing it to a sim that is the same size as the up-res initially. The up-res-method has a speed gain of about 30% from my test. The file is attached!

image.png.70935c6d9fa0388a83ceadb009419ecd.png

That's the compare after 100 frames. Geo 2 is the sim that is high-res initially.

Cheers,

Martin

simple_upres_bunker_comparison.hipnc

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...