Phoenix FD - Epic Smoke Test

[Paul Oblomov]

A lil test with upcoming new version Epic smoke made easy!

Full frame. No compression.

[bos]

Awesome Job. Makes me want to set aside more time to play with phoenix...

[LarsSonparsson]

wow! That looks really good. Any numbers to share?

[Paul Oblomov]

Grid size - 60millions. 2 hours to simulate 210frames. 5minutes per frame average on i2600k.

[G0M0]

How dense grid did you used?
I see you have some strange turbulence artifacts which I have also - the kind of axis shaped split on new born wave of smoke.
I still don't know how to reduce this kind of effect.

[Paul Oblomov]

This is symmetric conservation. Maybe this is the cause.

491x353x347

[G0M0]

I've got this issue in smooth conservation also.

Here is what I am talking about:


they are best visable moving to camera but sometimes it is clear visable on the sides

[Ivaylo Katev]

unlike the particle based simulations, the grids are anisotropic and such kind of artifacts are always present. the most used solution against such artifacts is adding turbulences/noise that hides them. especially for phoenix i can say that the advection method is very important for this problem. but there is no better advection, sometimes ft is better, sometimes the classic one

classic


ft


classic


ft

[G0M0]

Yeah, I thought it is something unavoidable :/.
But good to know there are ways to workaround.

[PIXELBOX_SRO]

was that rendered in vray with GI on or is it scanline???

[Paul Oblomov]

Bruteforce + lightcache. And 6 lights to guide vray.

[PIXELBOX_SRO]

hmm ....strange ...my frame times in Phoenix are quite brutal :-/ when it comes to fire...it might be smaller with smoke only as that itself (in your case) doent emit any light.

[Ivaylo Katev]

if you mean that the fire slows down the render speed, the image sampler is the reason. the fire is bright and this causes a lot of rays, you have to decrease the upper limit of the sampler. the fire itself is even faster than the smoke, because there are no shadow rays for the lighting.

[PIXELBOX_SRO]

good tip! thanks Ivaylo.
Will mean chenge of the workflow quite a bit cos i am using the sampler at 1/16 or 1/8 quite a bit to get rid of the noise which vray 2.0 tends to produce on whites.
Will have to explore this and possibly render the Phoenix stuff in separate passes or regions

[Paul Oblomov]

I think you can clamp and use subpixel mapping. Using 1/12 here.