How to model a light source using a spectral distribution curve for the color

Question

My previous attempt was unclear so I'm rephrasing the question:

I'm making a promo video for a certain type of lighting and at one point, I'd like to show the difference in the spectral power distribution (see graph below) of the light output for each type of bulb. In other words, I want to use the SPD to color the light emitted by the bulb into the room based on wavelength and output power.

I have the SPD, I just need to use it somehow to color the emission source. This is all I can think of but it's not exactly mathematical and I'm not sure it really captures a realistic representation and I don't know if it just basically averages the colors together:

It produces this light:

But I would rather have the emitted light follow a curve similar to this if at all possible:

Thanks for your help.

You may want to consider using an engine such as LuxRender or Maxwell which will directly support SPDs via a text file or the like. — JtheNinja, Jun 20 '17 at 02:12
That's very interesting, I've never used anything but Cycles, I'll look into it. — bertmoog, Jun 20 '17 at 02:17
I have used the Blackbody node but I didn't even know about the wavelength node... checking it out... — bertmoog, Jun 20 '17 at 03:17
Hmmm... would be neat to have an input to the wavelength node to determine the Q of the color selection but I'm not sure how to do it. — bertmoog, Jun 20 '17 at 03:23

score 3 · Answer 1 · answered Jun 20 '17 at 05:13

3

To accomplish this you would require a wavelength to XYZ conversion, and then from XYZ to sRGB / REC.709 primaries conversion.

Make sure that you obtain linearized RGB values from your wavelengths. If you are using normalized wavelengths, display linear (0.0 to 1.0) would suffice.

answered Jun 20 '17 at 05:13

troy_s

12,268
2
36
75

1.) When you say XYZ to sRGB conversion, do you just mean to ensure the display color space is in sRGB (i.e. there's no sRGB conversion node unless you mean combine RGB)?
2.) How will using the XYZ channels help to plot a curve over the spectrum with certain high Q peaks? – bertmoog Jun 20 '17 at 15:49
Also, according to this: https://blender.stackexchange.com/questions/31375/is-separate-rgb-identical-to-separate-xyz .... the RGB and XYZ nodes are identical except for nomenclature. – bertmoog Jun 20 '17 at 16:40
Uh... Not spatial XYZ. CIE 1931 XYZ. You won't be able to use Blender to achieve this. – troy_s Jun 21 '17 at 01:20
Ah, I see. Blender can't do this even for a simulation? – bertmoog Jun 21 '17 at 03:15
2

If you were careful and know the precise values to set up in nodes as well as the intended meaning and how to use the results, I suppose a very driven person could do it. It would be hell however. :) – troy_s Jun 22 '17 at 15:33
Ok, I appreciate the info – bertmoog Jun 22 '17 at 15:39

How to model a light source using a spectral distribution curve for the color

1 Answers1