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

anon123

That's interesting. I'm new to modern Macs as you can see (my last Mac that I truly used for significant period of time was the CRT iMac lol).

At this point I am a bit tired of chasing new monitors and hardware swaps. Focusing now on how to build up defenses (breaks, eye washes with water, acupresssure, etc. ) and using the screens less.

anon123 when macOS divides by exactly 2 it works by just doing a 200% zoom, the same as Windows, so the vibrating feeling they had is probably related to usual Mac temporal dithering. Sounds just like the dithering symptoms I get on iPhone SE 2, most Macs, AMD video cards etc.

however, yes there will be additional distortion and moire patterns on not-exactly-2 scale factors (this is where it does differ from Windows). But exactly 2 should be fine, so they're definitely just feeling the symptoms of dithering

JTL I think PM permissions are disabled by default. I had to manually ask admin to give me permissions for it.

So after some more experimentation with the microscope, it turns out the OLED display I was testing Stillcolor with seems to have built-in FRC dithering. In fact, it is quite aggressive. The marketing and independent specification for the display make no mention of this. It occurred to me that I might be able to confirm that there is no temporal dithering built into the display by looking at the onscreen controls under the microscope. Sure enough, the middle grey design elements in the onscreen display settings have temporal flicker.

With that, I am more confident that Stillcolor is working on the M1 MacBook Pro 16". And that the temporal dithering visible under microscope might be LCD inversion with a sub-pixel implementation. But it amazes me how bad the LCD inversion is. The built-in MacBook display performs well on the many inversion tests available, but under the microscope it is pretty much terrible. If this flicker is LCD inversion, then I could see LCD inversion being a main cause of people's discomfort with these machines. I didn't know it was this bad, and had always thought this sub-pixel independent flicker under microscope on darker colors was temporal dithering.

Stillcolor does for sure increase the banding on the Lagom gradient test. So there is evidence it is working. It would be awesome to see time blend output from an M1 MacBook Pro 16" to a capture card, but for now I'll just have to assume Stillcolor is working. Because of the very similar noise of inversion, it is so far not obvious that anything temporal has been disabled by Stillcolor, through the microscope.

In the meantime, I am going to try to come up with an experiment to distinguish between LCD inversion and temporal dithering (sub-pixel type) to put this confusion completely to rest.

Blooey i still think the XDR Pros are dithering with an additional 10-bit to 8-bit FRC step like external monitors do (because macOS output is 10-bit but I doubt the internal panel is "true" 10-bit), and this would best be solved if Stillcolor gains the ability to truly force color depth

can you try measuring the display with both Stillcolor enabled and BetterDisplay dummy/virtual display mirrored to the physical display? BetterDisplay dummies try to render 8-bit (but I'm pretty sure the actual output of the Mac in the end is still 10-bit, until progress on forcing bit depths is made), curious if the flicker pattern changes with that configuration

DisplaysShouldNotBeTVs @aiaf in addition to the obvious change made by disabling uniformity2D DISABLING "NormalModeEnable" ALSO MAKES A HUGE CHANGE

( FYI, when disabling NormalModeEnable, make sure that ProMotion is off (use 60 FPS or lower.) Even though it also affects ProMotion, trackpad mouse cursor movement will get totally messed up, but this isn't an issue on 60 FPS and lower.)

after disabling:

at default gamma the screen will slightly change color tint and dark grays will get brighter

at darker gamma the screen gets A LOT brighter overall, and background colors look more solid

disabling it also makes shadows on windows feel less 3D (maybe because the level of backlight zones now becomes more similar between a dark window and a lighter background?)

AND on some pages, like the animation when you load the "MacBook Pro" page on Apple.com, there is even more banding after setting NormalModeEnable to false. (look at the white gaps between the laptops that are animating)

again, disabling this simply makes the screen BETTER, it removes another one of the strange tricks used to mask the imperfections(?) of the display, potentially this one is related to control of the mini-LED backlight zones?

this image on Apple's website implies that typically mini-LED zones will have different brightness per zone that attempts to match whatever image is being shown?

potentially this property reduces or disables this?

there are still some blotches of (now not dithered anymore) gray on white backgrounds visible at lower gammas, but we're just a property away from clean screen output (aside from forcing 8-bit which is also important), i can feel it

So far:

enableDither = false: way less text shimmer, can more easily visually process multiple occurrences of a repeating object at once, edges of "pixel-perfect" icons in Finder list view look sharper

uniformity2D = false: the "vignette effect" / fade at edges is GONE

NormalModeEnable = false: less "cloudy", more natural color tint

both uniformity2D and this potentially reduce the weird "fake 3D" effect

The issues that remain:

"Fuzzy text/glowing halo around text" (may be related to panels from one of the multiple manufacturers only, based on prior experience since one friend's Mac didn't have this issue), some shimmer on certain background colors, blotches of slightly different shades on solid backgrounds (that are very obvious now because they can't "blend in" by being dithered)

@aiaf do you have any pointers for being able to use IOKit in Swift on iOS so I could experiment with adjusting enableDither on my iOS devices? IOKit is a private framework on iOS and gives an error when trying to import it with IOKit.

Some people said they could use it through Objective-C instead, but I don't know how to write ObjC at all (despite me being very experienced with Swift). However, even when trying to do this just to see if it would work, I actually couldn't get IOKit to load in ObjC either.

———

FYI: As I am a mobile developer, I have a LOT of iOS devices I've collected over the years… so I can certainly help with testing anything iOS-related if this ever becomes something you're interested in.

For reference:

Devices I own with "good screens": iPhone 4 (iOS 7), iPhone 5 (iOS 6.1.3 native, but also runs iOS 7/8 via dual boot), iPhone 6 (iOS 12), iPhone 7 (iOS 15), iPad mini 2 (iOS 10.2), iPad 6 (iOS 15)

Devices I own with "bad screens": iPhone 5s (iOS 12), iPad Pro 11" 2018 (iOS 17), iPhone SE 2 (iOS 17.2.1), iPhone 14 Pro (iOS 16.4)

DisplaysShouldNotBeTVs Update: I'd recommend only disabling dithering and uniformity2D — as disabling NormalModeEnable seems to make the system less stable / cause the cursor to lag / eventually randomly crash the WindowServer and log out.

Disabling the other two, dithering and uniformity2D, works totally fine and that does most of the work in improving the screen, so disabling normal mode isn't that necessary.

I'm still not sure which property will get rid of the "blotches of slightly different colors".

I wonder why disabling NormalModeEnable is causing this?

@aiaf just found AN IMPORTANT ONE

Disable VUCEnable

This causes even more banding to appear on the gradient test than just disabling dithering on my M1 Max XDR.

Additionally, "irregular patterns of boxes" appear on certain solid color background shades.

Potentially this is disabling the internal panel's FRC???

If you turn the Mac backlight up to max and Software Brightness (BetterDisplay) very far down, you will also now see a "dark, heavily banded blotch that follows the mouse cursor like a trail" when moving the mouse on a dark gray background. (Easily seen after zooming in with Ctrl+Scroll accessibility zoom.)

This confirms another thing I swear I could see for years, where it always felt like there was a dark spot around the mouse or that there was an area flickering around the mouse (specifically on the internal display). There IS, and it was literally being dithered (i.e. flickered) to hide it, too.

This will also make a strange "animation" visible (that was present before but now very obvious) of all the "blotches" I've pointed out moving around whenever Software Brightness is changed, lagging behind a second or so with what seems like an intentionally smoothed out animation — even when BetterDisplay's own "software brightness smoothing" is disabled.

I still don't know how to disable these blotches, but one thing's clear — they're basically programmed to "move around in the corner of your eye at a slower, smoothed out rate compared to everything else" whenever colors change on screen. And before enableDither=false, uniformity2D=false and VUCEnable=false, the effect was being dithered very aggressively in order to "hide" whatever this is.

tmfd Go to Stillcolor at the top left of the Xcode file tree, then "Stillcolor" under TARGETS, then Signing and Capabilities. If you see stuff like "App Sandbox" and "Hardened Runtime" here, there should be trash can icons at the top right of each of them. Click the trash can on both so all you're left with is one tab called " Signing". This will remove the app sandbox and allow those properties to be changed.

Then, just comment out the alert.runModal() line that shows the alert box that says "Stillcolor Issue". It will say "invalid argument" some times but then end with a "successful" anyway in the console after it tries a couple of times, so even though it makes it "look like" it didn't work, it does make the change as long as app sandbox is disabled and you eventually see "returned 0 -> (os/kern) successful" in the Xcode console.

For example, here's what I get in the console when I disable VUCEnable:

(Disabled = box checked in the Stillcolor menu bar BTW)

AppleCLCD2 service: 310b
AppleCLCD2 client: a17b
IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
Failed to set VUCEnable to false IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
AppleCLCD2 service: 310f
AppleCLCD2 client: a17f
IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
Failed to set VUCEnable to false IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
AppleCLCD2 service: 3113
AppleCLCD2 client: a183
IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
Failed to set VUCEnable to false IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
AppleCLCD2 service: 3117
AppleCLCD2 client: a187
IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
Failed to set VUCEnable to false IORegistryEntrySetCFProperty returned -536870206 -> (iokit/common) invalid argument
AppleCLCD2 service: 311b
AppleCLCD2 client: a18b
IORegistryEntrySetCFProperty returned 0 -> (os/kern) successful
VUCEnable set to false

It says "failed" a bunch of times, but eventually ends with successful, and the VUC change works in the end.