V-ray Next 4.30.02 - Extreme "dancing" white pixels on any material at a distance with glossy reflections

[Joelaff]

After further testing I would say, you are correct, Lele. Not a lot of difference in the highglights-- if you want them small. With some GGX tail adjustment (say 1.5 ish) then they both look pretty decent, and also likely flicker less. The VRayLight as a disc (edit) without the custom texture map looks worse than the direct light to my eye. It has the 1990s phong look to it due to the harder edges of the hot spot.

The tespot doe not have zero diffuse or metalness of 1.0, mind you.

Direct Light:



VRayLight Disc with a falloff type VRaySoftBox map:



VRayLight with no falloff texture map:


I do like the ability to tweak the falloff and hot spot in the texture for the VRayLight, especially if we make the sun bigger, which I might do for something so reflective, though the Direct light has two advantages: It renders a tiny bit faster, and and means you don't have to figure out other lighting tricks to illuminate your entire ground if you see it (assuming your ground or other objects are spread far and wide and need the specular light).


Scene file here.
LIghtTesting.zip

[kosso_olli]

I still stand by the statement that it should produce correct highlights without flickering, regardless of any whataboutism regarding any other software.

And so do I!
Also, that the sun was never intended to use as is, but to exclude it from specular and reflection is completely new to me. And that kinda defeats it's purpose, especially in the field of automotive viz. Metallic and pearlescent car paints need the sun specular and reflection to look right, to get the right amount of flakes etc.

[^Lele^]

Also, that the sun was never intended to use as is, but to exclude it from specular and reflection is completely new to me.

As i mentioned, it used to be enough to hide the sun disk ("invisible" toggle) to also exclude it from creating speculars, behaviour that's been changed in time to the same split contribution as the other lights.
We were doing the hiding/custom sun thing 15 years ago, Olli, as the bread and butter of a lighter.
Trigonometry or algos haven't changed since, the sun is still tiny, still enormously energetic -> same results. (PGL currently won't help either.)

Regardless, if you can find a render engine able to achieve what you wish (i.e. not suppress, converge.), we'd be very interested in hearing out about it.
We're not being obstinate, it's just that there is no solution known to us yet, so if another engine manages it, the ideas may be worth exploring.

In any event, this is a topic of active, fairly high priority research, as is most of sampling, so while for now it's requiring workarounds, maybe it won't in the future. (NNs?)

[kosso_olli]

Well, thanks for the help and the detailed explanation, Lele.

We were doing the hiding/custom sun thing 15 years ago, Olli, as the bread and butter of a lighter.

But somehow I feel bad now.

[fixeighted]

I wouldn't feel bad.
It seems to me that we were presented with a tool that clearly wasn't and still isn't documented clearly enough.
If it was, then this particular situation would probably have been more widely known about.
It's a shame that info wasn't shared earlier is all.
I am absolutely sure there are many many more examples.

[^Lele^]

No need to feel bad, i wasn't leaning on you Olli, i was stating a matter of fact: the sun was always far too bright to use for actual speculars, so one used what would work.
The sun and sky documentation is quite old indeed in specific parts, and should be brought up to current spec (then one would know without testing that the sun disk isn't 300 float, but 1 million plus in the recent models...), ideally with samples and solutions to specific shading issues.

I'm making a point of bringing this forward with the docs team and helping out with.
Bear in mind we're in the middle of high summer, so people may be on holidays.