Weird Smoke Pattern

[vandy]

So I have this downward smoke stream and a cellular pattern appears, like it's textured and stays in place. Attached is the settings, and here's the link to the animation: https://www.youtube.com/watch?v=KbcF...r#action=share
Any ideas why this is happening and how to fix it? Much appreciated.

[Neilg]

Probably from the displacement. At a guess, i'd check on 'surface driven' and see if that helps.

[vandy]

Thanks Cubicle. It does look better than before (surface driven on), but the pattern is still there. Also, the light is a direct one on the left side, but there's some strange bright sections on the right side. This looks more like a rough than a final; any other ideas? https://www.youtube.com/watch?v=9StH...m-upload_owner

[vandy]

So, on the pattern, to fix it after checking surface driven, I upped the light subdivisions and it fixed it. http://www.youtube.com/watch?v=paKVF...ature=youtu.be
However, I can't seem to get rid of the white edges opposite the lighted source. Really weird.

[Ivaylo Katev]

Hi Vandy,
i would recommend you to avoid any fake upres tricks, especially the displacement, especially for animations. You can mislead an arbitrary observer with static image, but with animation there is no chance to achieve the same realism as the natural big resolution. And the quality is not the only reason. The displacement renders much slower than undisplaced content and the final result is that 100M grid simulates and renders faster than displaced 1M grid. Yes, the undisputed advantage of the artificial upres is the predictability. This is really important if you have only one day until the deadline, but if you have at least week, i would recommend to use big resolution.
Thats why in our last release all the samples have 10M+ grid resolution.

[Paul Oblomov]

He can easily use upreas with simple interpolation. It creates that crisp, hires feel. But 40million is a starting point for the upres to work nicely.

[vandy]

Thanks Cubicle, Ivo, and Paul. Besides the light subdivisions that helped with the shadow problem, and adding a second light to the right side, the thing that got rid of the white border on the right was under Colors and transparency\Transparency the Simple Smoke Factor was reduced from default .3 down to .04. About the only thing I might change now is increasing the Brute force subdivisions. It's not perfect, but this looks a lot better. Attached is the adjusted settings, and here is the animation link http://www.youtube.com/watch?v=RNQTB...ature=youtu.be

[Ivaylo Katev]

for only 5M cells that looks pretty good!
btw, about the GI i asked people who have big experience in the volumetric fx, and they were quite clear - it's better to spend the resources to increase the resolution and to have good motion blur (in case of explosions) instead to calculate GI.

[vandy]

When it finished simulating it actually grew to about 40M (adaptive grid). I don't have any explosions here, but what do you mean ..."instead to calculate GI."? Are you saying to turn GI off in Vray?

[Paul Oblomov]

40million Very nice.

[Ivaylo Katev]

When it finished simulating it actually grew to about 40M (adaptive grid). I don't have any explosions here, but what do you mean ..."instead to calculate GI."? Are you saying to turn GI off in Vray?

yes, exactly, turn off GI when possible. with no GI you can render 2-3 times bigger grid in the same time, and in the most cases it will look better than low res with GI.

[vandy]

Here's a new one with a close-up as well. Looks good, but at the bottom area where it's pushing through the air, there's flashing (see close-up). Is this a Vray setting or a simulation problem? Any idea how to fix it?

https://www.youtube.com/watch?v=p24OBvmeHm0
https://www.youtube.com/watch?v=lky7GA0lCVY (close-up)

[Paul Oblomov]

It is lightcache in phx. Never used it.

[vandy]

Thanks guys, looks how I wanted it. I tried several combinations with the LC in Phoenix and Vray GI: LC on/GI on, LC on/GI off, LC off/GI on, LC off/GI off; also using one light and two lights. Single frame renderings @1080p ranged from 4:33 to 26:00 (LC off/GI on). Turning LC off really increased the rendering time and made no difference; not sure what it helps with. So what finally worked was leaving LC on, turning GI in Vray off, and arranging my lights in the scene (brighter light has multi of 5, shadowed side much dimmer) to get the best look. Rendering time per frame ranged from :09 at the start and ended at 4:04. By the way, the simulation went up to 267M cells.

http://www.youtube.com/watch?v=oC42A...ature=youtu.be

[Paul Oblomov]

I have strange feeling with latest nightlies, that I just can't see the details. It looks to me, that you don't have ~300 million of cells. Very damn smooth I can't finish my volcano, because can't get desired look. 400million of clels looks like 100million.