Brute Force GI quality on very reflective, chrome surfaces

[tricky]

Following some really helpful threads on this forum, I am changing my methods for rendering my scenes. I am now using BF plus controling my samples per lightsource and material with a lot more 'respect'.

I'm working on a few scenes at the moment where most of the GI noise in the rawGI element is reasonably smooth, but on certain objects (those with a chrome material applied), the GI is very speckley indeed. On the towel rail on the left of this snapshot, I have adjusted the GI multiplier in vray properties, but on the other highlighted areas, you can see the speckles.

My BF is set to 128 subdivs. The subdivs on that chrome material are also set to 128, with a glssiness value of 0.98 and reflect bounces at 3.

Adaptive DMC is set to 1,4 with Clr threshold of 0.01.



Why is this happening, and is local multiplying of GI samples the best way to solve it?

[Morne]

I often get that on my billboard plain people, and they have NO reflection whatsoever

[lllab]

looks more like coplanar faces?

[joconnell]

Ah!

This isn't actually a problem at all. If you've got light rays in vray, they can either bounce off the surface as in diffuse light or GI, be reflected as a reflection or spec, or go through the object like a refraction. If you've got A really reflective or refractive surface, then they can't also be very diffuse. If you've got a material that's got a 50% reflection value, that means there's only 50% left for the diffuse part of it (that's the energy conservation law you hear about a lot - the total amount of light that comes back from a surface can't be higher than what's hitting it to start). If you've got a 90% reflective chrome surface, then there's only 10% of diffuse left for the raw light and raw GI passes to hit. What vray does is that if part of the shading doesn't contribute a lot to the FINAL pixels in the beauty, it won't spend time sampling it here.

What you've got to look at with the raw light pass for this case, is also the diffuse filter. Since diffuse filter x raw light = final lighting pass, if you've got a raw pass that looks really shitty in places, see if the same section in your diffuse filter pass is really dark. If the shitty area in the raw light is being multiplied by really dark values in the diffuse filter, it means that the grainy raw light areas are being knocked down hugely in the final result, so the grain won't be seen.

If you've got a 95% reflective mirror, it's got 5% diffuse which is a tiny amount. Vray won't spend much effort sampling this since it doesn't contribute a lot to the final image. Same with highly refractive things like glass - it's almost as if they don't have any surface for the light to fall on to, so vray won't bother spending any time worrying about their light and gi shading.

EDIT

Also lllab is right - it could well be the edges of the mirrors and glass sitting on the same place as the wall faces, move them slightly away from each other or put the glass through the wall and vray'll sort it out.

[tricky]

Ah!

This isn't actually a problem at all. If you've got light rays in vray, they can either bounce off the surface as in diffuse light or GI, be reflected as a reflection or spec, or go through the object like a refraction. If you've got A really reflective or refractive surface, then they can't also be very diffuse. If you've got a material that's got a 50% reflection value, that means there's only 50% left for the diffuse part of it (that's the energy conservation law you hear about a lot - the total amount of light that comes back from a surface can't be higher than what's hitting it to start). If you've got a 90% reflective chrome surface, then there's only 10% of diffuse left for the raw light and raw GI passes to hit. What vray does is that if part of the shading doesn't contribute a lot to the FINAL pixels in the beauty, it won't spend time sampling it here.

What you've got to look at with the raw light pass for this case, is also the diffuse filter. Since diffuse filter x raw light = final lighting pass, if you've got a raw pass that looks really shitty in places, see if the same section in your diffuse filter pass is really dark. If the shitty area in the raw light is being multiplied by really dark values in the diffuse filter, it means that the grainy raw light areas are being knocked down hugely in the final result, so the grain won't be seen.

If you've got a 95% reflective mirror, it's got 5% diffuse which is a tiny amount. Vray won't spend much effort sampling this since it doesn't contribute a lot to the final image. Same with highly refractive things like glass - it's almost as if they don't have any surface for the light to fall on to, so vray won't bother spending any time worrying about their light and gi shading.

That makes sense. So the RawGI is showing me, in effect, 'diffuse' samples. On a mainly reflective material, Vray assumes (rightly so) that you won't really see much of a diffuse value on that pixel - you are really seeing other surfaces reflected in it. So it only leaves a few samples (hence the speckles) in the rawGI for that surface. On a perfectly reflective mirror, the rawGI will therefore be black as there is, in essence, no diffuse at all on that object at all. Right?

[joconnell]

Bingo - it's the exact same as raw reflection - if you've got a surface that's only 5% reflective in the final image, the raw ref will look really trashy in that area. If you look at it in concert with the reflection filter pass (which shows you how much the raw ref is actually getting used by the brightness of the pass) you can tell that a noisy area might not actually be seen all that much in the final thing.

This is one of the speed benefits of vray - it won't sample things that aren't important via the threshold values.

[Morne]

So then in those cases its pointless to up the reflection samples to clean it up right?

[joconnell]

bingo - there's no point in making vray overcook everything for an area that won't be seen. You've got to use the raw passes in concert with the other elements to see how much the different areas are contributing to the final image. If an object is grainy in the raw lighting and it's diffuse filter colour is pure black then you've no worries. If you've got grainy raw light on an object and it's diffuse filter is a bright colour then that grain is going to show up. For the raw reflection channel specifically, you can also have a look at the normal reflection pass to get a quick idea if the grain in the raw is going to have any actual influence on the final look of the frame.