Noise/Grain error

[KSTAR]

I'm not sure if I'm dealing with a LW 2020 issue or a Turbulence FD issue on this one

I have a very simple scene using TurbulenceFD. The scene consist of a floor object (simple plane) 1 TFD container and 1 fire emitter that is sitting on the floor, no light sources other than the light emitting from the fire, and the environment is pure black. The problem is once I turn the roughness value above around 5% for the floor object no matter what value is in the reflection samples or camera AA samples I can't get rid of the orange grainy noise that is coming from the fire. The noise is evident when viewing the Specular Indirect buffer

Here is a list of things I've tried to fix it

-Camera and Reflection sample settings -multiple settings including extremely high over kill values
-Multiple material - BSDF, Delta, Conductor BSDF material with multiple PBR texture maps with various combinations of inputs being used or not used (normals bumps etc)
-AA adaptive sampling on and off with multiple values
-Limited Dynamic High Range on and off with multiple values
-Global Illumination on and off
-Lowering the fire temperature, and adjusting the dampening, and white point values in TFD all it does is produce the same noise/grain of the new fire color from the new settings
-Adjusting the fluid shader noise threshold in TFD
-Turning on indirect sampling in the volumetrics settings (does that even apply to legacy volumetrics? since that's what TFD works with)

Here's one of many render attempts this particular one is using BSDF material 25% roughness no 50% spec 0% metal no bump or normal maps all the other renders using different materials and settings still come out with the same type of noise

[prometheus]

use vpr various modes for checking direct illumination, indirect illumination, glossy surface etc, it filters out all other channels except selected..and from which channel here this noise is showing up, that is where you need to adress it.
I would guess glossy reflections in the surface is on, that is causing it, try turning it off the very first thing you do.

If it´s not that, you should go through vpr with each channel that may cause it.

[KSTAR]

Yeah I have already done all that glossy reflections is on and needs to stay on. The noise is only in the Specular Indirect channel because the only light source in the scene is the illumination coming from the TFD fire. The only solution that kind of works so far is rendering the Specular Indirect and Legacy Volume buffers and comping them in Fusion with the Rank Filter attached to the Specular Indirect image. It cleans it up a bit, but I'm trying address the issue natively in LW, and no combination of settings I have come up with so far will do so

[jwiede]

Yeah I have already done all that glossy reflections is on and needs to stay on. The noise is only in the Specular Indirect channel because the only light source in the scene is the illumination coming from the TFD fire. The only solution that kind of works so far is rendering the Specular Indirect and Legacy Volume buffers and comping them in Fusion with the Rank Filter attached to the Specular Indirect image. It cleans it up a bit, but I'm trying address the issue natively in LW, and no combination of settings I have come up with so far will do so

Is the noise also changing frame-to-frame (resulting in flicker)?

If so, that may be unavoidable in LW2020, there are some known issues with anim frame-to-frame instability.

[KSTAR]

Looks like there is no inherit solution in LW. Thanks to a tip from David Ridlen who was having the exact same issues no matter what settings he used, with Spaceship metal surfaces in space scenes. The problem is the absolute black environment we were both using. The only solution is to change the environment from black to grey and lower the gama in post. I made a post in the Facebook LW User Group with some before and after screen shots

https://www.facebook.com/groups/ligh...9732598132718/

[jwiede]

Looks like there is no inherit solution in LW. Thanks to a tip from David Ridlen who was having the exact same issues no matter what settings he used, with Spaceship metal surfaces in space scenes. The problem is the absolute black environment we were both using. The only solution is to change the environment from black to grey and lower the gama in post. I made a post in the Facebook LW User Group with some before and after screen shots

https://www.facebook.com/groups/ligh...9732598132718/

Ugh, that's disappointing. It's basically saying that LW2020 (really, LW2018+) aren't suitable for rendering many types of space shots, which were a long-time "bread and butter" VFX use case for Lightwave. Between that, and how volumetric lights now "work", that's likely to have a serious negative impact.

The problem with that workaround is the effect it has with starfield-type backgrounds: The level of gray environment needed to prevent noise also appears likely to substantially reduce the contrast between stars and background, affect shadows and their intensities, etc. Lowering frame gamma in post will have a significant impact on those elements, and while masking might help with volumetric lights, it isn't really practical for starfields (or shadows -- shadows are everywhere in those types of shots).

Yeesh, wish they'd told us it was to be an "LDR PBR surfacing/render engine" sooner, I'd never have upgraded past LW2015. All these proposed workarounds for noise/fireflies are just seriously reducing/limiting renders' potential dynamic range, in one way or another. Fine if you don't need it, but a deal-breaker for those who need HDR in renders.

Guess I'll need to run some tests with LW2019 and gauge the impact there.

[jwiede]

You know, thinking about it, it seems like most of the noise/firefly problems come down to a lack of a (user-controllable?) high-band luma filter applied to the samples during the roughness "microfaceting" for specular surfaces. Problem is, I don't think you can get at that efficiently/usefully via nodes (would love to be shown incorrect). I wonder if LWSDK would allow creating a custom material or even shader model to apply such a filter for those specific cases?

[KSTAR]

I just got a message from Kelly Meyers on Facebook he was able to get similar errors using illuminated hypervoxels. He mentioned it could be an issue with how adaptive sampling deals with gamma? I know that when I view the images in linear color space they're clean once I apply 2.2 gamma the noise is there. What's also interesting is that its right around 5% roughness where the noise just instantly pops on at 4% and below nothing, at least in the scene I was having issues with. 5% roughness in other scenarios isn't very high and doesn't product noise, but in absolute black I guess it does

[Thomas Leitner]

Hi,
it is difficult to find a solution for your problem without your scene setup.
If you do not use GI all noise in your scene must come from reflection. So, as always, find the needed value for your camera AA samples and then crank up your reflection samples. The noise will disapear.

I did a test with TurbulenceFD fire and a plane with BRDF material (25% roughness, 50% specular). Black BG is not a problem. Only reflection samples matter. Of course to save rendertimes you can try a noise filter (best: Open Image Denoise form DP Filter).

ciao
Thomas

[KSTAR]

Hi Thomas thanks for the reply.

The problem is specifically in the Specular Indirect channel which is corrected by reflection samples their is no noise any in other channel. I spent 2 days an every possible setting imaginable in LW and TFD I cranked samples for both camera and reflection to extremely high values and let it render overnight and it looked no different then when I set it to lower realistic samples. I turned GI on, GI off If you click the link in my earlier post you can see the details I posted on the LW Facebook Group page. One factor may be that my fire emitter is a write on effect that is being generated from a texture UV mapped to the floor object, where your fire emitter is separate and above the floor? I didn't try DP denoise because it cant be used for animation correct?

The good thing is I have an easier fix in post to fix the solution

[Thomas Leitner]

... One factor may be that my fire emitter is a write on effect that is being generated from a texture UV mapped to the floor object, where your fire emitter is separate and above the floor? ...

Yes, my emitter is a ball above the floor. I did an other test where I places the emitter directly on the floor.



Maybe you could post your scene and object files to find the underlying problem.

... I didn't try DP denoise because it cant be used for animation correct? ...

There is no guarantee that Open Image Denoiser works correct for animations, but we used it a lot for animations with out any problems (of course your base rendering should not be to noisy).

ciao
Thomas

[KSTAR]

Actually do you mind posting your scene since its so simple?

[Thomas Leitner]

Actually do you mind posting your scene since its so simple?

Here it is:

TFD_reflections.zip

You have to simulate the TFD simulation again.

ciao
Thomas

[tischbein3]

Ugh, that's disappointing. It's basically saying that LW2020 (really, LW2018+)

I would check if the scene renders correctly in 2019....also ran into some weird problems with 2020 when it comes to reflections...2019 seems to be more consistent.

[Thomas Leitner]

I would check if the scene renders correctly in 2019....also ran into some weird problems with 2020 when it comes to reflections...2019 seems to be more consistent.

I’d first check if the scene renders correctly in LW 2020. I never had any issue coming from black environment.