LuxCoreRender / BlendLuxCore

Blender Integration for LuxCore
GNU General Public License v3.0
718 stars 90 forks source link

Dispersion: ROYGBVI, is this order a bug, or the scene? #926

Open Sarah-C opened 1 month ago

Sarah-C commented 1 month ago

Is this some physical setup I don't understand?

It appears that we get:
red, orange, yellow, green, cyan, VIOLET, BLUE

Expected:
red, orange, yellow, green, cyan, blue, violet

.blend file: https://untamed.zone/bugReport.blend ( 179KB )

Steps to reproduce:
Make a prism, shine a slit through it, set the dispersion and IOR so the spectograph reallllllllly stretches out.
I've edited the Lux Core example scene by increasing the slit length and removing the "mist". The slit is angled down by a few degrees from horizontal. This has the effect of avoiding overlapping colors through the 3 dimensional aspect of mist, and limits the visual dispersion to the reflections from the white floor and wall.

Settings for the prism node used in the Blender file:
image

Here is the physical positioning of the slit, and the prism:
image
image

Screenshots:
image

Physical image reference:
image

A Flickr physical image with increased saturation:
https://www.flickr.com/photos/chris_malcolm/16145118188
image

Error message from Blender console: N/A

OS: Windows 10 Blender version: 4.1.1 LuxCore version: All of them (But used BlendLuxCore v2.9 alpha 1)


I'm aware the Lux Core Render site has a rainbow render in the example scenes (which I based my test render on). I went back to it to take a closer look of its output, and noticed the cyan/blue overlap looked suspiciously violet (when the color here should be tiny amount of green, and lots of blue.... as in bluer cyan), and on the deepest darkest blue part there wasn't any violet to be seen.

So for a clearer assessment, I pumped up the brightness and increasing the saturation a tad, it would appear the area blending from cyan to blue contains a suspicious amount of purple(violet) too, with a little violet seen again on the opposite side of the blue.
The eye's cones see the colors around 480nm as a bluer cyan (strong blue cone activation with less green cone activation), so there should be no red/violet in these areas.

This image suggests there's something about the violet in the rendering that's always been there.

https://luxcorerender.org/example-scenes/
image

Sarah-C commented 1 month ago

Appears to still be an issue: https://github.com/LuxCoreRender/LuxCore/issues/262

Sarah-C commented 1 month ago

Notes I've found on this issue: https://forums.luxcorerender.org/viewtopic.php?t=6564