[Solved] Burned out Refractions with secondary Gi solver

[Duron]

Hey folks,

i came across a problem which is annoying me a lot.

scene:
red glas with refraction IOR 1.1
behind that glass i got a override material - black gloss, reflect 1.0, reflect IOR 1.52

When i render with default settings (primary and second Gi set to Path tracing) the parts behind the glass looks like burnt out. I managed to find the causer and its the secondary GI solver.

Is this a bug?

note: take a look at the purple pixels in attachemnt 3.

[Ludvik Koutny]

You should never disable secondary solver. It's not problem in this case. What happens when you disable secondary solver is that no secondary reflections are being traced. That's why it looks different, but it does not look correct.

The thing here is that Corona is photorealistic renderer, so you need to watch your albedo. In nature, there are almost no perfectly transparent, or perfectly saturated materials. In this case, i have suspicion you use 100% red color. Try something like 220-240 saturation, but never 255. As for value, 255 may also not be a good idea. Try something like 240-250. That should make it look a lot better.

There's also no need to increase MSI to 64. In exterior car shots, raising MSI should make very little, if any difference. What will help though, for car shots, is to decrease GI/AA balance to 8, from 16, to focus sampling more towards AA. And if you are not doing any ultra high resolution prints, then you can also set internal res. multiplier to 2, which will help with Antialiasing around highlight areas.

[Ludvik Koutny]

I've made you also short video showing how refractive albedo influences appearance. The reason why you do not encounter this with other renderers is that Corona uses WideRGB color space, which looks better in some cases, and worse in others (like this one).

[Duron]

Oh wow, that's some support, thanks for the video!

I've noticed that too. I managed to fix this problem by decreasing the refraction from 1.0 to 0.5-0.75 but the solution you've mentioned seems more legit to me.

Here is an update - the purple pixels are gone. AWESOME

[Ludvik Koutny]

Yep, the thing is that with absolute red, and absolute (1.0) refraction, as much of incoming light is being reflected back into the camera in a red component as was received, so it will cause clipping. I usually just avoid using absolute values for anything. Once Corona 1.0 is out, it will be even easier to tweak it exactly to one's liking thanks to the interactive renderer.