About noise in general

[flino2004]

instead changing BF settings did you try tweaking LC.. i don't remember well but i thing if you play with lc color thre. would help you out with the noise.

Fernando

[pixela]

glad to hear you're getting close to solve this one, it does look like a tough one, but it's a beautiful scene already.
don't know if you tried that already, but maybe setting the glass objects to not receive and cast shadows may help with the noise as well. if it does you can have the best of both worlds, the glass and no noise

I tried blocking shadows during my tests but this didnt help at all.
I also tried it in various ways, excluding from sun or from object properties and from vray properties.
There is some strange behavior with the glass that I can't name.

[pixela]

Well, I seem to have resolved part of the problem. First of all, when rendering the scene without any glass I got a very noticeable reduction in noise. This doesn't help me, though, because I need all this glass here. I then removed all fog colour, which is a shame but something I can live with, and got the noise to hit an acceptable level that way. It also got rid of the strange shadow artefacts I was getting in RT.
Mostly, however, the problem seems to have come from my own approach. It seems I was following the wrong noise-busting strategy. As Rivoli suggested, I tried to lower the AA settings, both the threshold and the subdivs, and to put pretty much all of a burden on the Brute Force subdivs and noise threshold. By raising the subdivs to 60, I could get rid of 80 per cent of the noise and hit a relatively reasonable render time. It doesn't look like this would ever be a fast scene to render, but at least I can get it clean. I'll post some images of my tests in the next few days if anyone is interested.

This is very interesting. I wonder why.
Then is there a difference in render theory with Vray? I mean after reading articles about Adaptive Subdivision and the Universal Settings-AA parameters taking care of all the noise, now is interesting to hear that cranking up AA settings higher can't solve the noise but instead making BF subdivisons higher solves problem.

I wish Vlado or someone on Chaos team explain the behavior so we can understand the theory better.

[BBB3]

I think cranking up the AA settings would solve the noise eventually, but (for this scene at least) it seems they would have to be cranked up a lot, sending render times through the roof.
One other thing I observed while running those tests yesterday was that even minor increases in the AA quality settings seemed to be having a major impact on render time with relatively little payoff in terms of noise while I seem able to raise the BF subdivs pretty high, with great benefits on noise, without impacting render times in a huge way.

[pixela]

Thanks for the input. I will try this too.
But there are some things I don't understand... Making BF subdivisions higher is supposed to solve GI noise issues. But on the other hand, if I have noise problem caused by reflection glossy, it doesn't have much effect.
In your case, did this solve all noise problems?

[BBB3]

I didn't really have much noise in the glossies to start with. Indeed the biggest problem appeared in the white diffuse materials.

[pixela]

I see. Then it makes sense. If you don't have any reflection or glossy but mainly diffuse white, then probably it was GI noise that disappears with BF subdivisions.
I still can't understand the behavior of glass.

[PIXELBOX_SRO]

So how would you tackle the glossies noise? Turning the individual material subdivs up simply?

[peterguthrie]

white diffuse materials are usually the areas I have most problem with too.

Wondering if you can be bothered to uninstall vray2 and try a render with 1.5 just to get a straight comparison?

[BBB3]

So how would you tackle the glossies noise? Turning the individual material subdivs up simply?

I would have tackled them with aggressive AA settings but now I guess higher individual settings should be the way.

Peter: I'm not sure I can, to be honest Though I should.

[peterguthrie]

I probably could... just need to remember which scene I have problems with!

[pixela]

I probably could... just need to remember which scene I have problems with!

Please post your results if you can make this.
Did you too have problem regarding the glass?

[PIXELBOX_SRO]

Bertrandt: something like 1/16 for the DMC AA? or even bigger?

There is something which i dont understand with the new vray 2.0 sampling in the first place......if you consider the workflow:
LWF LUT setting / color mapping gamma 2.2 / dont affect colors on / sRGB pressed in VFB ......should not this give you washed out image???

i dont really use BF/LC cos it takes long ....i use IM/LC instead....but i always tend to finetune individual material by raising their subdivs instead of raising global parameters
BUT
i have one scene now on which i could try BF/LC and see if i bet any issues you guys talk about...loads of glass in too))

[pixela]

I didn't really have much noise in the glossies to start with. Indeed the biggest problem appeared in the white diffuse materials.

Just one idea pops up in my mind...
What about making camera exposure higher, so the image gets brighter which will make GI sampling easier and then saving the image 32bit and correct the exposure...
I wonder if this can solve the problem.
If yes, the source of the problem can be located.

[lukx]

Bertrandt: something like 1/16 for the DMC AA? or even bigger?

There is something which i dont understand with the new vray 2.0 sampling in the first place......if you consider the workflow:
LWF LUT setting / color mapping gamma 2.2 / dont affect colors on / sRGB pressed in VFB ......should not this give you washed out image???

i dont really use BF/LC cos it takes long ....i use IM/LC instead....but i always tend to finetune individual material by raising their subdivs instead of raising global parameters
BUT
i have one scene now on which i could try BF/LC and see if i bet any issues you guys talk about...loads of glass in too))

I think it's more 1/100 as stated in vray manual universal method.