difraction and additive color blending error

[ior]

See this image:

error

All emitters have the same intensity of 1 for each RGB value, and have the same distance from the wall.

The traced circular diffraction used (that is similar to the eye) ends in square shaped form, with a thin cross of light horizontal and vertical.
Is variable depending on the color of light source. Cyan and green have a dark spot in the place of the emitter, blue get to be magenta near the light source, magenta and red don't get lighter near the light source.


The color blending seem to have errors too. The yellow color made by green and red is not as bright as the pure yellow emitted by yellow light, but orange is a little bit yellowish even with a fainter yellow light, the same happen with cyan.
Magenta and red mix, only get brighter magenta, not reddish magenta.
Blue seem to have little impact on cyan magenta and yellow.
Other combination's of colors may have the same problems.


Does RGB 1 0 0 get the same light intensity as RGB 1 1 0 or RGB 1 1 1, I think it should.


If I am wrong in something just point it out.

[galinette]

The color blending seem to have errors too. The yellow color made by green and red is not as bright as the pure yellow emitted by yellow light, but orange is a little bit yellowish even with a fainter yellow light, the same happen with cyan.
Does RGB 1 0 0 get the same light intensity as RGB 1 1 0 or RGB 1 1 1, I think it should.

Hi,

How did you make your light sources? Did you set a blackbody emitter with a RGB value? Or did you do a custom spectrum?

Secondly, 1W red + 1W green + 1W blue does not make a white light in real life (I do not know in Indigo). That's much more complicated than this. Usually it will be purple.

Then, I'm not sure that in Indigo, when you define a 1W emitter by a blackbody and a RGB value, you end up with 1W light. That's only the case if there is no RGB value specified, or if the RGB is 111. I may be wrong though (testing required...)

RGB 1 0 0 does NOT have the same intensity as RGB 1 1 0 nor 1 1 1, for sure (as a simple rule you can assume that intensity Y = 1*R + 4.59*G + 0.06*B, but for the CIE 1931 2° space with D65 white balance, from linear RGB values)

Etienne

[CTZn]

I don't understand your rig fully but it's looking nice !

Dark spots are due to the Mitchell-Netravali filter values: reduce ring parameter or use another filter.

It is true that the currently enforced diffraction mode can show some limitations... there used to be two modes for diffraction, I'm missing the other one tbh.

Emitting at a power doesn't prevent the energy to be dispatched around; the same amount of energy over a bigger surface would look dimmer I suppose.

[dag]

You need to tune the rgb values in light layers to get the right source color. This can be hell since it's so sensitive.
The circular aperture shape isn't perfectly circular. Use the one in http://www.indigorenderer.com/forum/vie ... =19&t=9917
The square cutoff is either a bug or some top secret internal scaling algorithms. You get this if you have post process enabled with very bright lights.

[ior]

How did you make your light sources? Did you set a blackbody emitter with a RGB value? Or did you do a custom spectrum?

Just put a rgb value of 1 in emission to make each color.

Secondly, 1W red + 1W green + 1W blue does not make a white light in real life (I do not know in Indigo). That's much more complicated than this. Usually it will be purple.

Light power should be not measured in watts = electrical consumption (W = VA), because that depends on efficiency of the electrical emitter in certain wave lengths that depends on the type off emitter (I think). For Indigo it should be lumen per square meter or lux:
http://en.wikipedia.org/wiki/Lux
http://en.wikipedia.org/wiki/Lumen_%28unit%29

RGB 1 0 0 does NOT have the same intensity as RGB 1 1 0 nor 1 1 1, for sure (as a simple rule you can assume that intensity Y = 1*R + 4.59*G + 0.06*B, but for the CIE 1931 2° space with D65 white balance, from linear RGB values)

I totally forgot the white balance! but I tested with F11 white point an it does not change things much:

F11 blueish white point

And I think that if you use lux as measure unit, a light with RGB 1 1 1 or _RGB 1 0 0 should have the same intensity.

I don't understand your rig fully but it's looking nice !

The scene is composed of one wall and black tubes with a sphere emitter inside, all at the same distance to the wall so you can see the the emitter itself and it's projection in the wall. The camera is centered.

Dark spots are due to the Mitchell-Netravali filter values: reduce ring parameter or use another filter.

I did not use that filter and it seems that it does the same thing with other filters but Ill try to change filter type and parameters.

Emitting at a power doesn't prevent the energy to be dispatched around; the same amount of energy over a bigger surface would look dimmer I suppose.

If the emitters have all the same size and distance to the surface and are not reflected or refracted they will all have the same intensity.

The circular aperture shape isn't perfectly circular. Use the one in http://www.indigorenderer.com/forum/vie ... =19&t=9917

I'll try that. But in your image there are still the magenta diffraction on blue and in all colors the diffraction is different, I think they all should be equal, starting to be white at the center and then fading out with emitter color.

[CTZn]

I don't understand your rig fully but it's looking nice !

The scene is composed of one wall and black tubes with a sphere emitter inside, all at the same distance to the wall so you can see the the emitter itself and it's projection in the wall. The camera is centered.

Dark spots are due to the Mitchell-Netravali filter values: reduce ring parameter or use another filter.

I did not use that filter and it seems that it does the same thing with other filters but Ill try to change filter type and parameters.

Emitting at a power doesn't prevent the energy to be dispatched around; the same amount of energy over a bigger surface would look dimmer I suppose.

If the emitters have all the same size and distance to the surface and are not reflected or refracted they will all have the same intensity.

I understand the rig now, thanks. The ring effect was much similar to the M-N artifacts, that's weird.

Tell me, are you using Reinhard tonemapping ? Linear is better for these colour tests I'm guessing, as the default scheme is somewhat tinting images. See how close to white you can get with just playing with the value. If you are using Linear tonemapping already, well, I'm off the topic

[galinette]

Light power should be not measured in watts = electrical consumption (W = VA), because that depends on efficiency of the electrical emitter in certain wave lengths that depends on the type off emitter (I think). For Indigo it should be lumen per square meter or lux:
http://en.wikipedia.org/wiki/Lux
http://en.wikipedia.org/wiki/Lumen_%28unit%29

Nope. Emitter power should be measured in candela, which is homogeneous to W. Lux is the measure of the amount of light received by a surface.

And I think that if you use lux as measure unit, a light with RGB 1 1 1 or _RGB 1 0 0 should have the same intensity.

That's wrong. The CIE RGB being a linear colorimetric space, it's additive. Then RGB(111) = RGB(100)+RGB(010)+RGB(001) and Power(111) = Power(100)+Power(010)+Power(001). Believe me or not, I do colorimetry and spectral optics as a living.

[ior]

Nope. Emitter power should be measured in candela, which is homogeneous to W. Lux is the measure of the amount of light received by a surface.

Ok I was wrong, but it was only a supposition. Thanks for clearing that out.

And I think that if you use lux as measure unit, a light with RGB 1 1 1 or _RGB 1 0 0 should have the same intensity.

That's wrong. The CIE RGB being a linear colorimetric space, it's additive. Then RGB(111) = RGB(100)+RGB(010)+RGB(001) and Power(111) = Power(100)+Power(010)+Power(001). Believe me or not, I do colorimetry and spectral optics as a living.

But If you diffract white sun light by a prism you will see all colors at the same intensity.
So if you have a yellow light and a red light with the same candela intensity, they will have the same intensity?

I tried in the first renders to put RGB(1.5 1.5 0) RGB(1.5 0 1.5) RGB(0 1.5 1.5) and RGB(3 0 0) RGB(0 3 0) RGB(0 0 3) and the result is this:

Additive color blending RGB 3000000 and Cyan Magenta Yellow 1500000 Emission value. Linear Tonemapping

the yellow from colors blend is even lighter than direct yellow of course. But the orange is still yellowish.

Tell me, are you using Reinhard tonemapping ? Linear is better for these colour tests I'm guessing, as the default scheme is somewhat tinting images. See how close to white you can get with just playing with the value. If you are using Linear tonemapping already, well, I'm off the topic

Just tried linear tonemmaping (above image) and no changes.

[ior]

The Inverse of the previous image, now the CMY colors match the intensity of RGB additive colors.

additive color blending RGB 1500000 and CMY 3000000 Emission value.

Pure cyan is a little bit greenish in relation to blend cyan.