I disabled dithering on Apple silicon + Introducing Stillcolor macOS M1/M2/M3

madmozg

aiaf nice example. I turned off dark mode in MBA M3, and its a bit better to my eyes. I don't know why, but gray/black colors giving eye strain more than light colors. If i'm opening VS code with dark theme its causing some eye strain and slight nausea.

madmozg

DisplaysShouldNotBeTVs I think MBP is using 10bit display, without FRC, according to specs from Apple. I know that Apple XDR Display is true 10bit, thats also mentioned on their website. I know that their regular Apple studio display is using FRC the same as MBA.

Blooey

madmozg I think MBP is using 10bit display, without FRC

On my MacBook Pro I do see a change with Stillcolor in that it seems to reveal quantization artifacts. Waves of magenta and cyan. Seems they have kept some level of dithering in place to further boost the display beyond 10 bits and manage those waves. It still isn’t clear if this is temporal or spatial dithering.

Thanks for those screenshots. It really seems to follow that “support for X colors” is Apple’s marketing for FRC dithering.

aiaf

madmozg Nice catch! Additionally, looking at the Pro Display XDR White Paper, the specs match the M3 MBP almost word for word, brightness (MBP has higher SDR nits), color gamut, refresh rates (MBP additionally has ProMotion), contrast ratio, reference color modes. The only thing that doesn't show up in MBP marketing is the "true 10-bit" part, which can be a marketing trick so as to not cheapen the Pro Display XDR.

I've now learned about a new variable besides the DCP, which is the display's TCON (Timing Controller). Does it perform its own dithering? To what extent is it influenced by driver properties?

anon123

DisplaysShouldNotBeTVs

no it's a big difference

macOS scale down
Windows scale up

try to disable "retina" mode in macOS and tell me if the text is more gentle and easier to read despite being super blurry

madmozg

Blooey I also saw it with MBP 14', dont know why. Looks like software tried to send 8bit signal, but GPU of apple silicon still using dithering or something like that.

"Support billion of colors" words used by some manufacturers like LG SAMSUNG ASUS and etc. You can check their websites and official specifications. In monitors manuals they will show you what type of display, 8bit, 8bit+FRC or true 10 bit, its easy, but I think not that easy with Apple. I wonder if this specification type forced to have by US regulators or so? And why apple don't have this kind of described documentation what kind of display etc used in their devices..

For example asus specifications for their monitors:

NewDwarf

madmozg This is just a commercial trick 😉

Apple claims about 1 billion fake color achieved by temporal dithering but they never say these are the "fake colors".

async

Lot's of references to pixelcapture in the flags. Wouldn't capturing image directly from the framebuffer be enought to diff for dithering, if laptop screens are mostly dumb? I saw some old references to capturing pixel data on Google, but no idea how hard it would be to actually make a working solution here.

Somewhat on the side, but if you could capture pixel data properly it would be fairly easy to do some logic that gives a score to how much changes there are to the screen. You could even calculate in a way where it would ignore things if the same image is shifted slightly in one direction for the next frame, so it gets a high score if you drag or scroll with choppy movement. There are massive differences in how choppy things are in browsers as well, based on mouse used, how the drivers interact with the browser, if smooth scrolling behaves well with it, if something is locking the ProMotion VRR among other things.

aiaf

async don't know how feasible it is to capture data that's sent to the built-in display, perhaps the m1n1 DCP tracer can help here. We have to contend with the DCP and TCON (if that's a separate chip on notebooks). Can anyone shed light on Apple TCONs? We also don't know what pixelcapture means in the context of DCP either.

madmozg

I have a Lumix S5ii and this camera able to do 30 shots per second. Maybe I can somehow capture dithering on MBA? Like setting proper shutter speed. And after using ffmpeg to make a video and see any differences/pixel flickering. I don't have any micro lenses unfortunately, to make a really close photos to display 🙁

async

Did anyone experiment with the logging flags? I couldn't figure out where they end up. Don't thing they are written to disk automatically at least.

async

Found these as well. Not that ChatGPT descriptions can be trusted, but seemed promising.

enableDBMMode: This flag likely refers to a "Dynamic Backlight Management" mode. Dynamic Backlight Management (DBM) is a technology used to adjust the backlight intensity of a display based on the content being shown. This can enhance the viewer's experience by improving contrast and detail in dark areas of the image while also reducing power consumption. In a device's framebuffer, enabling DBM mode would allow the system to dynamically adjust backlight levels based on real-time screen content, optimizing for both display quality and energy efficiency.

enableBLMSloper: This flag seems to indicate an option to enable a "Backlight Luminance Modulation Slope" adjustment. This feature would allow the system to control the rate at which backlight luminance changes, possibly in response to changes in environmental lighting conditions or on-screen content. A sloper adjustment could fine-tune the backlight for smoother transitions or more accurate representation of content brightness, enhancing user comfort and potentially contributing to energy savings by avoiding overuse of high backlight levels.

DisplaysShouldNotBeTVs

aiaf

The TCON also detects certain patterns that may be difficult for the LCD panel to display, and optimizes the display at the pixel level to minimize artifacts.

Sounds suspicious… deeper than the DCP level, "minimize artifacts", what is this optimization?

Hunter20

@aiaf Can confirm it works on my Macbook air M2 15inch (Ventura 13.5)! Thank you!

How to check:
Go to this page and enable/disable dithering while viewing test image --> http://www.lagom.nl/lcd-test/gradient.php
I'm clearly seeing banding appearing in the test image when dithering is disabled

My other findings:
ARM64 version shows colors in a different way compared to X64 (LibreWolf)
How did I test? Run this youtube video in LibreWolf ARM64-version and also X64-version, when compared, ARM64 looks more static to me. I guess this has to do with Rosetta.
https://www.youtube.com/watch?v=ksyYkA59LtU

Also very interesting, as @aiaf stated: "If you play a 60fps video on a 60Hz display while dithering is enabled, you actually get no dithering in the video for the most part. There’s no time to dither."
(Thanks to @async, you shared this github link before with us) https://github.com/abinabdc/flickeringMacFix
The program runs a small invisible video on the menu bar, the author of the program says it fixed flickering for him. Seeing your observation, maybe this can also be a different workaround with the same effect. I guess the program also forces the ProMotion refresh rate to stay always at 120hz. I have no ProMotion, as I have the Macbook Air, but maybe it's useful for people who have the Macbook Pro.

aiaf

Hunter20 this is a super interesting concept! But as noted in my original post, only the video portion of the display does not experience dithering, surrounding elements will still be subject to dithering.

However, we can perhaps overlay a refresh-rate matched video over the entire screen. This video will contain precise noise to subvert any subsequent dithering algorithms. However care should be taken so that this video does not become another source of eyestrain.

async

Would be interesting if someone with a microscope experiments a bit with Pro Camera. You can set custom fps and adjust a ton of options, as well as get a visual indication as to how much noise there is.

Also noticed that while zooming in you get this clear red and blue bleeding on white letters. Similar to what is annoying me.

photon78s

Lumix S5ii can do 24p/180fps? Perhaps trying capturing high speed video with the microscope held in front of the sensor (have to do this very carefully).

Has anyone noticed changes in pixel level flicker over time? In the morning my screen has less activity than when left on for a few hours later sampling with the scope at the same gradient dark area. Is this related to time/temperature compensation/timing controller?

https://www.cnet.com/reviews/apple-pro-display-xdr-preview/
In this article, they mention the pro display xdr maintain color accuracy over the whole day as opposed to other monitors which can do so only for 30 minutes!

Does the system do more/different dithering over time as an attempt to maintain color reproduction over time?

async

photon78s Absolutely. Text can suddenly start to glow again. I've been somewhat uncertain about if this is from the display, something closer to migraine switching on, or related to contact lens dryness / light scattering. Tbh I don't know. I do get similar effects on other text at times, but I've also had many cases where I reopen the laptop lid or change something and it seems to come on. The are multiple flags related to temperature and similar, so I guess in theory it could be. I've also usually felt that the screen is better in low power mode. Even when comparing with fixed refresh rate. I did not however check what framebuffer flags change when in that mode.

One slightly more far out theory of mine is that what is known but slightly changed triggers some other processing on the brain. Truetone is absolute hell obviously, but I often end up with way more issues after testing some color adjustments or resolutions. Or it could just be the reticular activating system and hypervigilance after focusing on it changes are good for a while.

Sure would be nice to have a straight forward objective measure for when problems are present.

Maxx

What would be the best way to get feedback on how many people have found resolution with stillcolor?

And how many of those are such that get red eyes from PWM and dithering,

karut

Maxx a survey tool online? Google offers one

aiaf

DisplaysShouldNotBeTVs I think I've figured out what Normal Mode is- it's the counterpart to Self-Refresh Mode (or Low Power Mode). Useful when the image is static and there's no need to still transmit 60 frames per second, so we save energy overall.

See
https://www.anandtech.com/show/7208/understanding-panel-self-refresh
https://hardwaresecrets.com/introducing-the-panel-self-refresh-technology/
https://patents.google.com/patent/US9640146B2/en
https://patents.google.com/patent/US8749541B2/en

This can explain the apparent lag with the cursor. Perhaps self-refresh mode pauses dithering too. Repeats only a single frame.
One experiment we can try is to toggle Normal Mode on and off with a timer at t intervals, starting with 1 second and see how that affects image quality, subjective eye strain, cursor lag, and system stability.

async

aiaf I have an election app that overlays live static noise. The type people here would panic when seeing. I've used it plenty when using the built in display. It's super comfortable at times. Almost like white noise for the eyes. Especially when I start to get annoyed by glowing text or focus problems. It locks the refresh rate obviously, but my main feel has been that it gives enough input to keep something in the vision system activated enough to avoid some issue. I also get the feel that it is easier to focus at the right distance to where the screen is. This however makes all pixels change continously, so I would guess it might also prevent some other stuff by accident. Also it keeps the gpu at super high load and eats the battery.

aiaf

aiaf I made a quick (dis)proof of concept of a dither busting overlay

https://gist.github.com/aiaf/8006286d0b38fe02518bf7f8a8890918

Plays a transparent checkerboard pattern every other frame over most of the screen at 60fps. I checked with my capture card, it does not alter the pixels in a significant way frame-to-frame for the DCP dithering algorithm to change its behavior. In fact, this overlay becomes another source of eyestrain and screen instability in its current incarnation. But maybe someone can experiment with different patterns, frequencies, or blending modes and see if that makes a difference.

Easier to just disable DCP dithering entirely with Stillcolor.

madmozg

I talked to a guy who works as an hardware engineer for very big company with their mobiles phones, I told him my problem with the displays, and he was shocked and hear about this for the first time. He mentioned that they are just getting the displays from other manufacturers like samsung/lg/boe and don't give a sh1t what that display is, how it works and etc. They just assemble it and sell, thats it. So maybe this info would be some how helpful 😃

« Previous Page Next Page »