sizzling reflection noise?

[jasonhuang1115]

Hi,

I have a circular piece of metal in the scene (animated rotating) that has a copper-like material assigned. The environment is an interior space with spherical HDRI projected onto the geo to through a VRayLightMtl to serve as the environment reflection.

With the material and render setting below, I got sizzling reflection noise in the rendered sequence.
No light, no GI.
material:
Diffuse Color: almost black (0, 0, 0.01 in HSV)
Reflection Glossiness: 0.8
Reflection subdivs: 128
Use Fresnel: On
Fresnel IOR: 20
Max depth: 3

Render setting:
ADMC sampler Min: 1, Max: 8
AA filter: gaussian, size: 1.5
Threshold: 0.005

DMC sampler
Time Dependent: off
Adaptive amount: 0.85
Adaptive Threshold: 0.01
Adaptive Min Samples: 8
Subdivs Mult: 1

Render time is already longer than I can afford (about 26 minutes in my i7 3.4 GHz 12 GB RAM) considering I am only rendering this piece of metal and the environment.

If I swap out the HDRI with a pure white VRay dome light, the reflection looks much cleaner/smoother, but the specular pass has little black dots sizzling across frames (very subtle though)???

Been fighting this noise for several days trying to get rid of it with the render time I can afford for rendering the entire scene. All comments are welcome.

Thanks.

[Morbid Angel]

you don't need to have threshold of 0.005 in your scene. Did you make sure you have no duplicate geometry there, or some issue with normals? Your settings otherwise look fine to me, I may suggest trying 3/16 for sampling it may help but if the sizzling effect stays there may be another issue.
If you want you can post a part of your scene to have a look at.

[jasonhuang1115]

you don't need to have threshold of 0.005 in your scene. Did you make sure you have no duplicate geometry there, or some issue with normals? Your settings otherwise look fine to me, I may suggest trying 3/16 for sampling it may help but if the sizzling effect stays there may be another issue.
If you want you can post a part of your scene to have a look at.

Thanks for Chiming in, Dmitry. I double-checked the normals and made sure there is no overlapping geos on those pieces of metals. It's interesting that as I raised the Min sampling (2 or 3), it didn't affect the sizzling noise a bit; but as I raised the Max sampling from 8 to 16, 24, and 48, the sizzling nose became less and less noticeable without sacrificing render time (telling from single frame). I am test rendering a sequence with 1/48 sampling and 0.01 threshold (ADMC Image Sampler) and see how it turns out.
I still couldn't wrap my head around this Max sampling thing. Sometimes, I had to crank up the number (ex: 1/48 in this case) to have better smoother glossy reflection; and sometimes I found a sweet spot of like 1/16 or 1/25 that reached the best result with the least render time? If you are tuning the ADMC (Image sampler) setting for a scene, would you start from like 1/4 and go from there to crank up the Max sampling?

[Morbid Angel]

I may be wrong but when you use 1/4, vray limits how many sample can be applied even if in your shader you set them to a high value. That's why you need to raise the max to make sure that you use higher sample value. In terms of render times, since dmc sample can mutate rays to handle all kinds of subdivisions, it takes longer to trace the rays if you don't have enough samples in your shader/light.

[jasonhuang1115]

I may be wrong but when you use 1/4, vray limits how many sample can be applied even if in your shader you set them to a high value. That's why you need to raise the max to make sure that you use higher sample value. In terms of render times, since dmc sample can mutate rays to handle all kinds of subdivisions, it takes longer to trace the rays if you don't have enough samples in your shader/light.

It makes more sense to me now after your explanation. I will check out the rendered sequence later tonight and maybe get back to you on this.

Thanks!

[jasonhuang1115]

I tested with 4 more different render settings. (all
Settings that are the same across all four test scenes:
no GI
128 subdivs for shader glossiness and rect lights
ADMC subdivs: 1/48 and Threshold of 0.01

Setting that are changed:
DMC sampler
Adaptive amount: 0.85 to 0.7
Adaptive Min Samples: 8 to 16
subdivs mult: 1.0 to 2.0

The sizzling noise is still there and doesn't become less noticeable at all as I raise samples in the DMC sampler. The only obvious difference is that the more samples I sent to the scene the better the direct lighting effect is represented, though with longer render time.

I am wondering if I should lower the quality setting in DMC sampler and up the ADMC quality setting by lowering the Threshold to 0.007 or 0.005 to maybe get rid of sizzling noise with similar render time?

[Zdravko]

You can try setting your HDRI with VRayDomeLight instead. Make sure DomeLight subdivs are sufficient.
Also to keep track of the sampling you can use Sample Rate render element.
The sampling is represented with a gradient from Blue>Green>Red. Red color shows the places where V-Ray set maximum amount of samples.

[jasonhuang1115]

You can try setting your HDRI with VRayDomeLight instead. Make sure DomeLight subdivs are sufficient.
Also to keep track of the sampling you can use Sample Rate render element.
The sampling is represented with a gradient from Blue>Green>Red. Red color shows the places where V-Ray set maximum amount of samples.

Hi Zdravko,

I would like to stick with the HDR-textured geo as the environment as I want to get spatially correct reflection from the environment. I thought about rendering a spherical image from the position of the CG asset and then plugging the rendered spherical map to a dome light for more efficient sampling or control but this approach is still less desirable.

I kept tweaking the Max ADMC sample and material glossy subdivs value and found that with the following setting, the render sequence in terms of sizzling noise.
ADMC:
1/50
0.005 threshold
256 glossy subdivs for the metal material
32 subdivs for all 6 rect lights in the scene

DMC sampler: all default

BUT, the render time is really unacceptable to me after a test rendering of the full scene (it finished about 31% after 8 hours of overnight rendering for the entire scene).

I noticed that with ADMC setting like 2/6 and material glossy subdivs of 36, though the Sample Rate element shows there are more samples (red) on the problematic metal piece, the resulting image is interestingly worse with bright tiny spots on the metal. As I raise the max ADMC sample and material glossy subdivs, the Sample Rate element shows greener on the metal (meaning less samples?) and the render looks much better with smoother glossy reflection.

Is there a way to optimize environment reflection for this case where the room is textured via a VRayLightMtl that has HDR map (bright lights in it) plugged to its color?

[mr furious]

it could be that from pixel to pixel you could be going from a really high value (ex. 100 to a lower value . in your HDR. This happens to me often as I shoot really high range HDRs and those really high values could be next to a really low value in your reflection causing a huge range for the AA filter to do it's job.

i will check the subpixel mapping button in the color tab along with clamp of 3 (for example as long as it's above 1). My understanding the vray will clamp any values over 3 (if that's what you set) and then try and AA it. It should do a better job with AA from a ratio of 3 from one pixel to .8 of the one next to it, versus the unclamped unsubpixel mapped example of 100 to .8?

make sense? that might help the sizzle an any other sparkles...

A

[jasonhuang1115]

I replaced the HDR-textured room geo (my source of environment reflection ) with a VRayDomeLight, thinking the built-in importance-sampling can help on this issue. The reflection RE is smoother and acceptable even with lower quality setting and less rendering time. BUT NOW, I am experiencing sizzling black noise in the specular pass??? I already turned on subpixel mapping and clamp output of 5 for all the tests from the very first post.

ADMC: 1/50 Threshold: 0.01
DMC sampler: default

DomeLight Subdivs: 16
All 5 rect light subdivs: 32

How would you suggest to get rid of sizzling black noise in the specular RE?
Also, if there is any way to optimize glossy reflection for the VRayLightMtl that has HDR env. map projected for env. reflection, I am all ears.

[Svetlozar Draganov]

Could you please send your scene to vraymaya@chaosgroup.com and we will investigate what cause this black noise!

[dangerweenie]

I'm having a similar issue in my scene. Did you guys find a solution for this?

Best,

Matt

[jasonhuang1115]

You might want to refer to this thread for noise cleaning: http://www.chaosgroup.com/forums/vbu...846#post550846

I have much improved result after discussing with the support team and going through the stuff and tests mentioned in that thread.