Issue with noise that won't go away with samples

[ssjenforcer191191191]

I have a noise pattern on the same surface that changes with each subsequent frame even though nothing in the scene is changing.

You can see on the side of the cabinet in the middle of the image how the reflection highlights change between frames.

I tried using default samples render settings, then i turned up the diffuse and specular to 3, but that just moved the highlight noise around but didnt get rid of it pretty much at all.

i increased camera samples to 4, and that didnt help, again just moving it around.  Rendering the same settings a second time puts the noise in the exact sample pattern, so is it noise or something else?

It is creating a distracting animation in the noise that won't get better with increased samples in the camera, diffuse, or specular.

My light sources are just quad lights inside this building, and I even turned off the arnold sky so it doesn't add noise through the clear windows.

Is there anything else I can do to avoid this without having to increase my render times too much?

[ssjenforcer191191191]

I am not at my computer, but I know there is a point_light inside a glass lamp object on the side of the cabinet with the issue.

Could this be my problem?

I have that glass material as default mostly with transmission. 

Is there something about a point light that would cause that?

EDIT:
It actually happens in more parts of the scene than just that.  In fact, it happens inside an elevator that is enclosed and lit only by a quad light in the ceiling.

Is this noise in the reflection highlights?  Shouldn't that go away with increased samples?

Are these fireflies?
I will try enabling adaptive sampling and see if that helps.

[ssjenforcer191191191 · Accepted solution]

It seems like adaptive sampling is fixing it.  The render time is a lot higher, so I will need to try and optimize the settings.

[ssjenforcer191191191]

[]

[lee_griggs]

They could be fireflies caused by the intense highlights from the metal base of the chair? What happens if you replace the material or increase specular_roughness for the metal material?

Removing noise tips:

https://help.autodesk.com/view/ARNOL/ENU/?guid=arnold_for_cinema_4d_ci_Tutorials_ci_Removing_Noise_h...

[ssjenforcer191191191]

It could be fireflies.  It seems to occur on many of the specular surfaces, including glass in the scene.  Mostly metal is being affected.

I am fine-tuning the Adaptive sampling values to get ideal render times.

1 hr renders get perfect without sparkles animating, but I am trying to get it down to 30 without much compromise.

Images:

the fireflies are very much dimmer and barely noticeable now from frame to frame.

The highlights on surfaces and glassware were the main spots it occurred. 

I didn't expect the render time to increase so much.
I will experiment with lowering the Max AA sampling or increasing the threshold a bit more.

[ssjenforcer191191191]

I think I have much better render times (down 50%) by increasing the threshold from 0.025 to 0.035, which goes from 50 min to 35 min.

I tried decreasing the Max AA samples, but that just reduced it by 4 min. 

[lee_griggs]

I can see why those super bright reflections would be the cause of the fireflies. Are you able to adjust the lighting/material to avoid it? Are you using ACES Color Management? It would help a lot with these sort of issues.

[ssjenforcer191191191]

I use all aces already.

The lighting shouldn't really be an issue.  It's quad lights and a point light.

The materials seem fine.  It's all kitbash secret lab stuff.

I used nodes to make them rougher already.

[thiago.ize]

Small lights are more difficult to properly sample from highly reflective surfaces and so will require higher AA. You can either increase material roughness or make the lights bigger.

Maybe it's just one of these lights that is causing an issue? Try rendering with only each light to see if you can identify the problematic light which you can then tweak (make bigger, make it invisible in reflections, etc)

[ssjenforcer191191191]

Ok, I didn't know that about lights.

I have only a few small lights.  The quad light in the room is 2 meters wide with intensity 1000 and default resolution 512 and 1 sample.  Do the resolution and sample of an individual light make much difference?

Could bump also cause this, or is the lights themselves more likely it?

[ssjenforcer191191191]

I'm trying to learn from this in the documentation:

"If the issue is noise in a specular highlight, you will need to confirm that the source is the direct light and not a secondary ray type (such as specular). This is easy to achieve by turning off global illumination; set the diffuse_ray_depth and specular_ray_depth to zero (this essentially turns off all global illumination). If the noise is still there, we know it is the specular component of the illumination model. If the issue is shadow noise, then we can simply toggle ignore shadows in the Arnold render settings, and the noise will completely resolve."

I don't see an option in render settings to "toggle ignore shadows" anywhere.

I also tried increasing the samples of my quad light from 1 to 4, but it does absolutely nothing.

I set ray depth for diffuse and specular both to 0 and it makes the majority of the sparkles go away, so it is indirect noise?

However, there is no samples option in render settings for Indirect Specular, so how do I fix this?

EDIT:
I tried halving the intensity of my quad light, and it had no effect.

I even tried turning off adaptive sampling and turning up my camera samples to 4 and the specular samples to 6, and this did not improve the highlights.  The same result when I increased diffuse samples to 6 instead.

Specular samples to 10 didn't improve either.

[ssjenforcer191191191 · Accepted solution]

So I figured out that the highlights are being caused by Rounded Corners node in my materials.

I ruled out the bump nodes themselves and found it was purely from the rounded corners I was using to make the objects look more realistic without hard edges and corners.

Some objects have materials with bump only that cause this issue as well.

Lowering the radius for the Rounded Corners (and increasing the samples just to be safe) seems to fix that issue.

Lowering the bump height helps too.

[lee_griggs]

Glad you found a solution. I'd be surprised if it was just the round_corners shader doing it as I have never seen that before. If you have a simplified scene that can prove that, I would be interested to see it.

[ssjenforcer191191191]

I have some images in my previous post that if you see the elevator in the foreground, the bottom edge loses the highlight issue when I removed only rounded corners.

The radius value ranged from 0.8 cm to 1.0 cm for most materials that I used it on.  I didn't really notice the bad highlights in the IPR until I started rendering for the first time.

I made it better by reducing the radius value for the node.

[ssjenforcer191191191]

I found that I still had some fireflies even with high Adaptive sampling.  I figured out that using clamping fixes it.

So possibly my quad light had too high of intensity for its size.  I would try to use more light sources with lower intensity in the future.