PDA

View Full Version : LW 2018.0.3 - Caustics increases light? Please cross check.



madno
04-28-2018, 02:12 AM
Hi,

have strange effects with caustics on / off in GI (LW 2018.0.3).
If caustics is enabled the scene gets more light.
The more window glass layers I attach the more the light increases.

See below:
Example room with two layers of window glass - once rendered with Caustics off once with caustics on.


Windows Layers = on
Caustics in GI = off
141443
Caustics-off_WindowLayer-1and2-on.jpg


Windows Layers = on
Caustics in GI = on
141444
Caustics-on_WindowLayer-1and2-on.jpg

Anybody like to crosscheck?

Scene:
141445

jwiede
04-29-2018, 11:36 AM
Well, I can confirm the same behavior appears to occur on Mac LW2018.0.3. Activating caustics in the scene greatly increases the amount of light in the "room" for some unclear reason, with more "trans layers" increasing light present.

As you have content demonstrating the issue, definitely worth reporting it. Hopefully Newtek can fix it quickly as well.

madno
04-29-2018, 02:44 PM
Thanks for cross check.
Sent report to Newtek.

MSherak
05-01-2018, 12:48 PM
Hi,

have strange effects with caustics on / off in GI (LW 2018.0.3).
If caustics is enabled the scene gets more light.
The more window glass layers I attach the more the light increases.

See below:
Example room with two layers of window glass - once rendered with Caustics off once with caustics on.


This is correct behavior since in the scene light is shining through glass. The IOR of the material is bending and the caustics are being added to the GI. Problem is the caustics in 2018 are thrown through the interpolation softness and smeared all over. So the effect just looks like boosted lighting. Turn of Interp mode and use brute force MC and the lighting will not jump as high since it is not being smeared.

Moved the scene over to 2015.3 to show the look a little better. I did not set the caustics settings to high so one could see the samples being thrown around. (using my BSSRDF shader in 2015.3)

madno
05-02-2018, 07:59 AM
Thanks for the input. I had the idea as well before posting. But, I think caustics could only increase the light, if the glass is a magnifier. A flat window bends the rays but does not concentrate them, right?. So, I thought a room with windows and caustics active could not get more light than the room without.

Nevetheless, it is interesting that 2015 shows the same effect.

To me it looks like LW is treating caustics as an independent lighting that gets added on top of the normal lighting instead of being integrated into it.

jwiede
05-02-2018, 12:39 PM
This is correct behavior since in the scene light is shining through glass. The IOR of the material is bending and the caustics are being added to the GI. Problem is the caustics in 2018 are thrown through the interpolation softness and smeared all over. So the effect just looks like boosted lighting. Turn of Interp mode and use brute force MC and the lighting will not jump as high since it is not being smeared.

If the window glass were concentrating light from a larger surface than just the window opening, that would make more sense, but the window opening is the same size as the window glass (certainly nowhere near the scale of difference as exists between the light levels). Refraction from an opening-sized piece of glass versus opening alone doesn't explain such a dramatic overall increase in light level. The light level increase is pervasive as well, so it isn't as if some areas' levels are being raised while others' are being lowered by refractive focusing either. Where is all the extra light the window is somehow gathering originating given the window offers minimal additional surface area versus opening itself?

There's also the matter of additional layers of glass visibly increasing the light level further, despite contributing minuscule amounts of additional light-gathering surface (at edges) per additional layer. Again, refraction alone doesn't seem to explain where all the extra energy is originating in these cases.