Pyro upres advise.

[sebkaine]

Hi Guys , 

 

i am struggling to understand the upres workflow in H.

i am trying to replicate a Marey Type Wind Tunnel.

 

http://www.musee-orsay.fr/fr/evenements/expositions/archives/presentation-detaillee/browse/16/article/mouvements-de-lair-etienne-jules-marey-1830-1904-photographe-des-fluides-4216.html

http://www.graphicine.com/wp-content/uploads/2014/11/%C3%89tienne-Jules-Marey_unterschiedlicher_Form_1900_oder_1901.jpg

 

The thing is that i can't use to much disturbance / turbulence the brute force way, cause i need to keep a smooth elegant flow for my fluids.

I just want to get a lot of detail when i have collision.

 

I have a 500.000 voxel sims that do exactly what i want. Then coming from a maya fluid background i said to myself , now piece of cake let's do 

some uprez to get those nice swirly detail that you get when you start to enter the 500*500*500 arena.

 

Unfortunatly when i go to 1Mega Voxel / 2  / 3 or 4 i did get some extra detail but it's quite the same sims, i don't have the extra high swirly detail i want.

So i would like to know why this logic of upresing doesn't work ?

 

I must be making something wrong but i would like to know what ? How would you solve that in an elegant way in H ...

i would feel guilty to go back to the good old Maya fluids on this one !

 

Cheers

 

E 

 

Ps : i need to replicate this in 3D not only with a 2D fluids

Edited February 14, 2016 by sebkaine

[Guest tar]

try volumeresampleSop to make the sim smoother, post sim. Then you could try a gasturbulenceDop on the resampled Sop, brought into a new sim. 

[sebkaine]

Many thanks for your help Marty , i never use the volume resample good addition to my toolbag.

The thing is that it smooth the sims but at the cost of loosing very fine detail.

i will also try a gasturbulence.

 

But the thing is, and my post title is not very clear , is that i don't want a post - process at the moment, i just want to understand how you work efficiently to go higher in rez with Pyro.

In maya i often work at 250*250*250 to setup my general motion then i know that from that if i go to 500*500*500 or sometimes even higher , i will gain a lot of very fine detail in my sims.

 

I am quite sure i can get the exact same thing in H but , i lake understanding on this subject. I have now a 250*250*250 sims in H that is perfect for what i need. 

But if i go to 500*500*500 i have nearly the same motion without the very high detail i want.

 

Maybe i am doing thing wrong , maybe the best workflow is to get a good sims and then use some post process + shading tricks ?

I need to get this quality 

 

the fluid is extremely smooth but at collision you have LOT of very rich detail !

i have the exact same sim that work nice a 250*250*250 but now i want those extra detail , and going higher doesn't give me this extra quality ...

 

So any advise on how to reach that goal would be useful !

 

Cheers 

 

E

[Atom]

I recently re-visted Peter Quint's tutorials on pyro and there is some good information in there about quality. He basically says with H15 there is not really a need to up-rez pyro, that was just a hack before we had better tools. I assume you are using the combustion model. What shape options do you have activated?

 

I would say a lot of the quality is done by the shader, not the sim.

Edited February 12, 2016 by Atom

[sebkaine]

Thanks for your answer Atom, basically i just use density / temperature / velocity, i don't think i need any combustion here.

Maya wise i would only use density / velocity + high detail solve + swirl so i don't know if temperature is crucial.

 

The big difference is that in Maya the main force that push your fluid forward is buoyency, so you can control you sims without a temperature grid,

and just activate it to add detail in your density. I have try to replicate this in H , it does work but at the end forcing the velocity grid with Field or POP velocity

is maybe not a good idea ...

 

The big difference with H is that your density behavior is basically dictate at 90% by your temp grid so the answer is probably in a better understanding on how to precisly control

temperature ? i will make some test with POP emitting temperature to move the fluid instead of POP directly emitting velocities.

 

For the quality at shader time would you have any great videos / infos about this ? cause coming from a maya fluid background it's pretty new for me !

In maya if you want quality you had to put your CPU on his knees simple as that ... how would you get very high detail swirl in post process ?

 

rest / rest 2 field + a curl noise at rendertime ?

 

Cheers

 

E

Edited February 12, 2016 by sebkaine

[Atom]

so the answer is probably in a better understanding on how to precisly control temperature

That is what the combustion model is about.

 

how would you get very high detail swirl in post process ?

That is what the shape options help you achieve. Dissipation, Shredding, Turbulence all combine as weights to modify the temperature field.

 

I still recommend watching Peter Quint's videos on the subject. They cover the combustion model in detail.

 

I too am still learning about this topic so I don't have any golden examples yet...

[3iart]

I would use density and temperature only leave turb and disturbance out. That is what is going on in the video. Yes there is atmosphere, but the bulk of the velocity is coming from extreme temperature interaction. The collision object should contribute temp as well. then play with the various temp settings in the pyro solver simulation settings.

 

I've had bad luck with the uprezer. I would just sim at high rez.

Edited February 12, 2016 by 3iart

[paulh]

Have you tried using your collision objects as a control field for your turbulence? That will keep your sim nice and smooth until the smoke touches the collision geo.

And maybe try playing with the density ramp to sharpen details.

 

 

ph_wispy_smoke_1.0.hip

[sebkaine]

Thanks a lot for your very helpful answers guys !

 

My biggest difficulty was how to push the density grid forward in a smooth and elegant way, while making very very detail turbulent swirl when collision occur.

Paul your setup is exactly the piece of the puzzle that i was missing thanks a lot, i didn't know this trick of control field. I would have use a POP sims to mimic this in maya.

 

3iart you confirm that temperature is definitly the tool to use to drive the density motion.

 

At first i try to use only velocity + density

- i inject a straight velocity field in my grid with a volume source

- i inject a velocity curl noise around the collision object

the setup was looking like what Paul do , but instead of working on the temperature field , i was directly working on the velocity grid.

i did get very accurate stuff for the global behavior but the collisions were not working well , and i'm sure the fact that i directly work on the vel grid was the source of the pb !

It's a maya habit , but i think it's a bad one cause you are kinda hacking what happen in real life ... and temperature variation is in fact the main element ...

it looks that in Houdini Pyro solver you mainly control the velocity behavior with the temperature grid only ? am i correct ?

 

Paul you use a static solver instead of a source collision object ? is there any real motivation for that ? does it give better result ?

from my test i tend to use @collision @collisionvel with a source volume ?

 

I would also be curious to know about what atom say about post processing in the shader , is there any great tricks i am missing ?

i agree that playing with the opacity ramp will be a key factor , but it's the exact same story in maya.

 

Do you have any advise or render tricks in mind ?

 

Thanks again for your great help guys !

Edited February 13, 2016 by sebkaine

[rayman]

As far as I know you don't need temperature at all. You can emulate buoyancy using any field you want. Just multiply vector field by any scalar field and add it to vel. This way your buoyancy is not restricted to just one direction.

[deniz]

it looks to me similar to vortex sheeding, like in the karma vortex street exemple, but guess to use a object as collision field and create a vectorfield to drive the direction and the curl should work

 

https://en.wikipedia.org/wiki/Reynolds_number

Edited February 13, 2016 by deniz

[sebkaine]

@pavel 

Thanks for your answer !

I definitly need to get a better understanding of how the pyro solve work in depth , cause in fact Pyro look extremely flexible and you can plug and do about anything you want ...

 

@deniz 

Thanks for the link and i agree for the global setup :

- 1 field to dive the directional vel

- 1 collision field that create the very detail curl

 

My doubt where on where to work ?

a- directly on the velocity grid with a density / velocity container

b- on the temperature grid with a density / velocity / temperature container

 

i couldn't say why but i think that option b (paul hip) is the most elegant way because :

- you don't hack directly the velocities

- you have all the extra controlers that comes by default with the temperature (shredding / turbulence / disturbance )

- you have a temperature grid if you want to play at shader time

 

Then i'm sure you can do whatever you want in a custom way , but my expertise in Pyro is not yet at this level !

 

Here are the 2 most interesting watch i have find on pyro (you have english subtitle in the 2nd one):

 

Edit :

i also add this artice by intel in 18 part which is impressively informative :

https://software.intel.com/en-us/articles/fluid-simulation-for-video-games-part-1/

 

if you have any other good ressources feel free to share !

 

Cheers 

 

E

Edited February 14, 2016 by sebkaine

[sekow]

what you could do too is utilize vortex particles on/around the collision geometry.

which would allowed for more direct able swirls, as you do not rely on random noise (turbulence).

[sebkaine]

Thanks for your input Sebastian ! when you said Vortex Particles what do you mean exactly ?

- are you refering to a houdini feature that have fly under my radar ?

- or do you mean implementing by yourself the principle of Vortons by advecting POP that inject vorticles forces in the vel grid ?

 

https://benedikt-bitterli.me/vorton-fluid.html

https://software.intel.com/en-us/articles/fluid-simulation-for-video-games-part-2

[sekow]

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

basically you direct where and in which direction the curls should go.. 

worth a try.

[sebkaine]

Thanks a lot, it definitly has flew under my radar ...

[Shurka]

I think you can add extra detail on render if you will use gridles advection or SELMA. You can read about it there : http://magnuswrenninge.com/content/pubs/ProductionVolumeRenderingSystems2011.pdf

Edited February 16, 2016 by Shurka

[sebkaine]

Thanks for your help Shurka !

But Gridles Advection + SELMA are for the moment to complex for my H level.

 

I will try to focus on understanding the solver in depth first ! 

[paulh]

Here's a nice thread on gridless advection with some examples:

 

http://forums.odforce.net/topic/16415-gridless-advection/page-2

 

Jerry Tessendorf even has the last word!

[Guest tar]

I reckon the most fail safe way is to simulate initially at your final res... think of lower res as a different take in a shoot. Never the same!