Pink Plane Lights

[dj_buckley]

So I'm just running some basic test with the Corona Sky compared with a just a plane light.

And I've noticed plane lights are tinted pink.

If you look at the image below the left side is Corona Sky/Portal and the right is Plane Light.

Both were set to 3% Noise Limit.

Tonemapping settings in frame buffer were the same for both setups (WB was at 6500).

The plane light was to set to Kelvin 6500.

The CoronaSky has been run through a Colour Correction Node with saturation set to -100.

The original idea behind the test was to see if I could get similar results to a plane at the window, using just the sky/portals and adjusting horizon blur/ground colour, and subsequently which would be quicker

With the two setups, I'd expect them to both be a shade of grey as they have an grey override material on.

But ... the plane lights are tinting pink even though everything is set to 6500 6500.

Why is this?  I tested it by opening them both in Photoshop and cracking saturation up to 100.

Also, although the results differ slightly, light levels are roughly the same, light falloffs etc look a bit different.  Why would the plane light be so much quicker? 3 times quicker?  And why are they pink?

I actually prefer the lighting from the Corona Sky/Portal method as it gives a nicer contrast, better falloff from light/dark and generally more definition to everything.  But does the difference really warrant a 3 fold increase in render time?  I'd rather suffer the slight drop in quality than such a huge render time increase.  I can only imagine this increases dramatically on a complex scene.

But first things first, why the magenta tint?

I should add, on numerous scenes in the past, lit with plane lights, I've found myself battling a magenta hue to renders, and always blamed my monitor or WB ....

[pokoy]

This came up before and the official answer was that according to the CIE specification (or whatever else it was), 6500K have a slight shift towards yellow/orange (extreme saturation in PS turns it pink). Can't find the post right now, will look for it and post back here if found.

Found it, the answer is in this post by Ondra:
https://forum.corona-renderer.com/index.php?topic=15026.msg95934#msg95934

[dj_buckley]

Even so, setting the light to 6500K and the WB to 6500K should return white/grey though right as they cancel each other out?

I'm assuming the problem lies in the Plane Light as the WB is set to 6500K for both setups and the desaturated Corona Sky works fine as expected.

[pokoy]

I guess 6500K for WB means 'no change' as in completely neutral. I guess that setting WB to 6500*(1.438/1.4388) = 6496,385... would neutralize the slight color shift in 6500K (which is  ~6503.6 K in reality). I took the formula from the Wikipedia section linked in Ondra's response.

[dj_buckley]

Although it's not neutralizing it :)

I'll comment on the other thread

[pokoy]

What happens if you use a Color Map set to 6500K nested in a Corona Output map and desaturated as you did with the Sky map in the light's color texture slot? Same pink tone?

[dj_buckley]

Nope, you get the fully neutral grey when sampling pixels.

Corona Colour on it's own - Pink Tint

Corona Colour in Output map - Pink Tint

Corona Colour in Output map desaturated - NO Pink Tint

The change/tint is so obvious in Interactive to just by changing the Saturation in the Output Map

[agentdark45]

Interesting! I often find myself knocking out some red of the curves in the VFB in scenes with lots of interior lighting. I guess this explains why!

[maru]

Well, here is my test with a plane light. :D

dj_buckley, can I have your original scene? (see uploader in my signature)

[dj_buckley]

Just uploaded it, all you need to do to see the effect, is select the plane light, then switch from Direct Input to Kelvin Value whilst in IR

[maru]

Thanks, I'll take a look.

[TomG]

The Saturation of 0.5 in the VFB does greatly exaggerate colors - with a lower setting, the light appears white. With that same setting, using Direct Input with a color of 255,254,255 also results in a noticeable pink (that is, even a tiny shift in color will be greatly amplified with a Saturation of 0.5 in the VFB).

[dj_buckley]

I set the sat to 0.5 on purpose to make it obvious.  When set to 0 there is still a noticeable to change when you just switch between direct input and kelvin.

[agentdark45]

I set the sat to 0.5 on purpose to make it obvious.  When set to 0 there is still a noticeable to change when you just switch between direct input and kelvin.

Another thing is using kelvin values for spot lights e.t.c comes out a lot "redder" than real life. For example a disk light set to 3200k in corona has a noticeable red/dirty burnt orange hue to it compared to various bulbs around my house of similar kelvin values. Not sure if it's fair to make this comparison...but I assume kelvin values in Corona lights should mimic their IRL counterparts.

So far I'm getting better results either knocking out some red in the curves VFB window or using custom RGB light temperature values.

[TomG]

A screen grab and swap between the two with Sat 0 does show there is a difference, ty! For me the question is, "How come Kelvin is never pure white?" as there doesn't seem to be any value that gives a totally neutral color. Whether this is expected / intended, I don't know, I'll leave that for the developers :) But it does seem that there is no value that is completely neutral in color (using the inverse Color Tint seems to be the only way to neutralize that).