V-Ray 3.50.01 - PhoenixFD fire causes noise and fireflies on reflective matte objects

[LudvikKoutny]

Hi,

yet another issue unfortunately. It seems that as long as a matte object that has reflection with imperfect glossiness (<1) reflects PhoenixFD fire, it renders as strong noise and fireflies, and does not converge into any usable result even after very long time.

This picture shows a simple PhoenixFD fire sim with two matte object teapots on each side. Right side teapot has perfect 1.0 glossiness, where as left side one has 0.5 Glossiness. While 1.0 glossiness converges pretty much immediately, 0.5 glossiness takes ages and looks more like fireflies than sampling noise. Something like trying to render caustics with unidirection path tracing


Repro scene is here: PHX_FireGlossyNoise.zip

Thank you in advance.

[georgi.zhekov]

Hello,

This is a known issue. You can take a look at this thread http://forums.chaosgroup.com/showthr...ding-artifacts

[LudvikKoutny]

Hi,

I am not sure I understand this completely. The thread mostly mentions problem with PhoenixFD lights, but as someone in the thread points out, the problem persists even when Emit Light in PhoenixFD Volumetric Rendering options is disabled, and that is my case too.

I can understand it being broken when emit light is enabled, as there is explicit sampling of some lights non-native to V-Ray, but when it is disabled, shouldn't PhoenixFD fire behave like regular self illuminated volumetric material that has no special kind of explicit sampling?

I may have reproduced the issue with VrayEnvironmentFog too. I've replaced the fire with a Sphere of VrayEnvironmentFog that doesn't have much of a density, but has very bright emission, and results are pretty much similar:


And scene is here:VR_VolumeGlossyNoise.zip

I still have some hopes these are two different issues in V-Ray, and that this one could be fixable inside of V-Ray, instead of Phoenix. It seems as if V-Ray somehow missed majority of reflection samples taken from volumes that have very low density and very bright emission. Not something related to if volume is explicitly sampled light source or not

[LudvikKoutny]

Hi, are there any news about this? Unfortunately this bug makes it impossible to ever render out glossy reflections of any phoenix simulation that includes fire, which makes correct integration of such effects onto live footage impossible.

It is not caused by phoenixFD lights. Even if I disable direct illumination on phoenix explosion, problem still persists, and as shown on the example above, it is not related to just phoenix, but rather any volumetric material of low density but high emission value.

Seems more like there is some problem with MIS sampling of volumes.

[Paul Oblomov]

Try this. Can't say it is realtime, but not ages. Me showing every window I changed.