Donux It is modulation depth, so 15006 times per second the brightness drops 79% when looking at a pure blue background. The question is why that doesn't happen with red. If this was simply the backlight flicker (not PWM) then at least for me it makes no sense that it doesn't affect red.
Could it be that KSF phosphor that is slower to respond causes this?
https://pcmonitors.info/articles/the-evolution-of-led-backlights/
Just thinking as I type here, and I'm probably butchering the explanation and understanding of the visual system. A lot of how the visual system works it is built around opponency of the following types.
- High / low luminance (black / white)
- Red / green
- Yellow / blue
Basically our vision detects a shitton of edges and gratengs thru these combos.
What actually happens if red doesn't flicker (due to KSF phosphor) while green that hit the M and L cones does?**
We have some rods that mainly respond to brightness (black and white), and we have 3 types of cones that maps to different colors. The type of color and input can make us use certain visual pathways more or less. So looking at large slowly moving black and white blobs will take a different path than looking at colored details.
Some quick info on the 3 pathways:
Magnocellular (M) pathway:
Processes motion, depth, and coarse visual details
More sensitive to low spatial frequencies and high temporal frequencies
Plays a role in detecting sudden changes or movement
Most vulnerable to flicker due to its high temporal resolution
Responds strongly to rapid changes in luminance
Sensitive to low spatial frequencies and high temporal frequencies
Studies have shown M pathway neurons can follow flicker rates up to 60-80 Hz
Parvocellular (P) pathway:
Processes fine detail and color information
More sensitive to high spatial frequencies and lower temporal frequencies
Important for form and object recognition
Less sensitive to flicker compared to M pathway
More responsive to steady-state stimuli
Can still be affected by flicker, especially at lower frequencies
Triggered by red-green opponency.
Koniocellular (K) pathway:
Less well-understood compared to M and P pathways
Involved in color processing, particularly blue-yellow distinctions
May play a role in eye movement control and visual attention
Less well-studied in terms of flicker sensitivity
Some evidence suggests it may play a role in processing rapid color changes
Maybe one of the problems of KSF phosphor is mainly present when there is high frequency flickering, as instead of a simple brightness dip that affects the Magnocellular pathway, you would keep hammering the Parvocellular pathway with edge / no-edge / edge / no-edge. 15000 times pr second.
Remember that white on the display consists of RGB. So to avoid this you would need to remove either all red resulting in a cyan image, or remove both green and blue resulting in a pure red image. This can be done with Gamma Control if someone wants to experiment.
One thing to note is that if making blacks brighter or color shifted this effect could be significantly increased, as pure black doesn't generate any signal for yellow-blue opponency. One could also argue that newer screens getting better with less bleeding and overly blue backlight could trigger less of the different types of opponencies that the visual system uses to detect edges.
Say you want to create the maximum amount of opponency signals for text you would make blacks slightly gray and purple (blue and red), and you would make the whites greenish yellow. The luminance contrast is there in any case. Compared to a perfect screen with pure black and white you would have 3 opponency channels instead of 1. Not necessarily beneficial, but worth thinking about.
In any case I'm a strong advocate for experimenting with small color shifts the make screens feel better, and I think the differerent flicker rates of the colors for the MacBook display potentially could be the reason it messes up people.
@DisplaysShouldNotBeTVs
13% 17:25:29 ➜ defaults -currentHost write -g NSCGSWindowSkylightSupportsMoreScaleFactors -bool yes