Dome light Light Select Renderelement changes color when GI is toggled...?

Hey guys,

i lately discovered something i don’t quite understand!
I have a simple setup:
A dome light connected to a light select render element set to type: diffuse. When i now toggle GI (Brute Force or Irradiance Map) on/off
the light select element changes quite a lot. This change does not happen with the light select set to type: raw.
I cannot quite figure out the reason why, since to my understanding the result should be the diffuse color multiplied with raw light. And both ingredients should not change based on GI.
Here is a screenshot:


Do you have subpixel mapping enabled?

Yes i do have subpixel mapping turned on. However turning it off doesn’t change the problem.
I did a bit of testing with the color mapping settings right now, but nothing changed the problem.
Here is a screenshot of my standard settings:

Untick linear color please. That is incorect and u should not tick it on. Also untick subpixel mapping for this test please. Can you also disable the clamping? - none of those thing should fix the issue - except for linear maybe…

The screenshot from 1st post. Is it light pass that are we looking at or beauty pass? Can u also disable ray distance please?

Hey DADAL,
as far as i’m informed those color mapping settings should say:
“Use a gamma of 2.2 for threshold testing on subpixel samples (more accurate) but dont bake the gamma into the pixels, instead keep them linear for comp. Also gamma correct sRGB textures should you find any.
Also clamp values at 5 so no breakout values in comp.”
That is usually what i want…but just to make sure i did the same test basically with factory default rendersettings, same issue:

Why is this an issue:
We have a comp pipeline where we automatically reconstruct the complete render from renderelements in post. We rely heavily on light selects with type set to “diffuse”. We use them
along with the diffuse color render element to recreate the raw light passes in comp, to grade on, since that seems to work better than using the raw passes directly.
At the moment i’m lighting a desert scene. WIth GI on, the scene is considerably darker than with GI off (?) and looks ok. With GI off the scene
becomes massively overbright, due to the dome light going nuts in brightness…i’m not at work right now, but i will post images later.


Humh I actually recently had the same issue in my scene. The GI pass was darkening light as opposite to non gi pass. I was using LC with store direct light so I assumed it was LC affecting my light pass so I just took it as it was. Didnt think it was a bug. Now you have the same think but you dont use LC so it makes me wonder whats going on here…

Here are some more images. The light setup is basically a vraySun (without sky) and a dome light for the environment light:



I think, when gi is enabled, the dome light’s contribution is divided. Its being split partially to gi rays partially to direct rays. So long as end result looks the same it does not matter. So, its normal that the dome light pass on its own is darker when gi is on since part of it is in gi.

So long as end result looks the same it does not matter.
If the contribution in the end would be the same, i wouldnt mind (although i still wouldnt understand why…), but as you can see my Beauty pass is significantly different.
With GI on, the Beauty pass is darker.
With GI off its overbright.

I think, when gi is enabled, the dome light’s contribution is divided. Its being split partially to gi rays partially to direct rays. So long as end result looks the same it does not matter. So, its normal that the dome light pass on its own is darker when gi is on sice part of it is in gi.
Here is an image of the GI and Lightselect added together, which i suppose should according to your information, match the result of a light select for a dome light without gi.
That does not seem to happen here for me:

Maybe a better/alternative Thread title would be:
Why does my beauty render become darker when i turn on GI?

I see, can you post part of your scene. I think something else in at play there, besides the dome light.

Actually I was able to reproduce what you have happening. I think since you use a light select for your dome, and not a raw light pass, the raw light select outputs the lighting energy without considering the gi portion, so it looks much brighter. I think what you need to do is measure the intensity of your raw light and subtract that difference from your light select. I got the result looking pretty close to the beauty, though not 100%.

I see, can you post part of your scene. I think something else in at play there, besides the dome light.
Here is a link to the scene…free from all references etc.
Actually I was able to reproduce what you have happening. I think since you use a light select for your dome, and not a raw light pass, the raw light select outputs the lighting energy without considering the gi portion, so it looks much brighter. I think what you need to do is measure the intensity of your raw light and subtract that difference from your light select. I got the result looking pretty close to the beauty, though not 100%.
Testing that…although i usually try to keep direct raw passes out of the equation, and recreate them instead in post, since they usually are pretty unclean and prone to problems (like not adding up correct to match the beauty).

Just as an inforrmation:
This issue does NOT affect a lightselect from any other light (like vray rect lights for example), they work flawless in the expected way…

Please see attached scene with nuke comp, the dome light is darker, however it reconstructs properly to match with the beauty.

Because of MIS (multiple importance sampling), when GI is enabled, part of the lighting is computed as direct lighting from the dome light, and part of it is picked up from the GI. When both are added together, you get the correct result. If your comps are set up correctly, as Dmitry pointed out, the render elements still add up perfectly to the beauty element.

V-Ray 3.0 has an option to disable MIS for area lights, including the dome light, but I don’t generally recommend doing that.

Best regards,
Vlado

Also regarding RAW passes. As far as I remember they come out a bit incorect from 2.xx version of vray and only 3.0 has it corrected - I might be wrong here.

If you are referring to the edge issues, this is correct yes.

Best regards,
Vlado

First: Thanks Dmitry for taking the time to test this…appreciate it! (Actually didnt know a sunlight could have a light select)

Because of MIS (multiple importance sampling), when GI is enabled, part of the lighting is computed as direct lighting from the dome light, and part of it is picked up from the GI. When both are added together, you get the correct result. If your comps are set up correctly, as Dmitry pointed out, the render elements still add up perfectly to the beauty element.
No doubt they are adding up to the beauty, that was not the issue for me. The issue was more of a workflow issue. When lookdeving i usually turn GI off, since its much faster without when you are tweaking stuff that just doesnt need it, like direct lighting, shaders etc. And you usually know what you get from GI, some subtle lighting on surfaces close to other surfaces. Its basically the expensive extra you just turn on at the very end, since it doesnt influence all the other work you carefully did before (Basically you turn it on before you send it on the farm…).
So i was pretty surprised when i saw that extreme change in the direct lighting (dome light) when toggling GI. To me, beeing able to turn that off in Vray 3 seems like a usefull option…

Would have loved to do the current production on Vray3. We had a lecturer here at Filmakademie just few days ago and he was full of praise for the version 3 (for Maya) :smile: