Using Eink Monitors To Detect Temporal Dithering

markdotpeters5

jordan I just did a test using OBS to capture my desktop, then the Ffmpeg script to highlight the changes between frames as linked earlier..

As a test I played a video on one half of the screen and had the lagom.nl dithering test picture on the other half…

Absolutely no movement on anything but the video, so looks like my 22H2 isn't dithering.

jordan

markdotpeters5 I dont think a screen record is going to be accurate

WhisperingWind

markdotpeters5

I agree with @jordan here, because software‑based screen recording usually pulls data from the OS framebuffer, while temporal dithering is added further down the pipeline, when the framebufferr’s contents are sent to the graphics card for processing and output. Therefore, frames from the framebuffer may not contain temporal dithering.

simplex

markdotpeters5 using OBS to capture my desktop

it does show nothing, because post-processing applied after OBS capturing frames stage in grapfic pipeline

markdotpeters5

jordan might be capturing before the dithering is applied or at a different level - have to do it with a capture card as per here: I disabled dithering on Apple silicon + Introducing Stillcolor macOS M1/M2/M3 - LEDStrain Forum I suppose

WhisperingWind

jordan

But does the system see the ultrastudio capture device as a monitor with a proper monitor edid?

The recorder is recognized as a TV. But the “monitor” I use with Windows 10 22H2 is also a TV 🙂

I hope more people can jump in with testing this.

I hope so, too, since it’s still unclear what the issue might be. I don’t think Windows 10 handles TVs and monitors separately; however, since my only true 8‑bit display is the TV, I can’t confirm or refute that.

markdotpeters5

WhisperingWind I've just tried with my capture card. It's limited to 1080p @ 60fps but clearly only shows any movement that is me moving the cursor/window around.. I'm not sure if it's capturing at the required detail though!

This is capturing on a Windows 11 laptop I happen to have here

WhisperingWind

markdotpeters5

If the capture card can record uncompressed video, the result will be accurate.

cooldudz

Man great discussion by Jordan and the community! Bravo everyone. Learning a lot reading this thread.

jordan

cooldudz
Thanks man!

So I have an update, guys… I loaded Windows 10 22H2 and tried what was mentioned above — disabled color calibration in Task Scheduler and a few other things — but nothing made a difference.

Then I went back to 1809, using the post-1809 wallpaper I had extracted, and set it as my wallpaper. It showed the same temporal dithering behavior. I even tried disabling wallpaper compression to see if that would help, but nope. When I set a solid color as the wallpaper, it just vanished — what the heck?

If I had a window not in fullscreen, the wallpaper would dither and cause the window to dither too. I’m not sure where else this might be happening. I checked Edge briefly and didn’t notice anything right away. I also tried another default Windows 10 wallpaper — and there it went again, even on 1809.

I honestly don’t know what Microsoft is doing. I’m tired of switching GPUs and testing different Windows builds, so I think I’m done investigating Windows for now — unless there’s something else worth trying. I really don’t like how Windows is doing this, and I don’t trust it, so I’ll be sticking with CachyOS.

edit: I did also try the intel arc a770 and quadro k4200 which didnt make a difference. I also did use the EDID spoof option in the quadro driver and set the edid from the up2720q true 10 bit monitor but that didnt help, it just caused things to dither like crazy.

markdotpeters5

jordan it's a frustrating myriad if options isn't it 🙁

I hadn't heard of CachyOS so I'll give that a try on my old Surface laptop gen 1

AshX

jordan Then I went back to 1809, using the post-1809 wallpaper I had extracted, and set it as my wallpaper. It showed the same temporal dithering behavior. I even tried disabling wallpaper compression to see if that would help, but nope. When I set a solid color as the wallpaper, it just vanished — what the heck?

Not a PC guy so I’m not sure if I’ll be much help here, but this behavior has been identified on Mac. I’ve combed through tons of threads on Apple Support, and while it won’t get as technically detailed as on this forum, there are more than a few posts from people who found the resolution of their symptoms by changing the wallpaper on a recently upgraded version of MacOS (most posts seemed to be about Big Sur to Monterey and Monterey to Ventura).

Since your methodology is relying on the E-Ink, it relies on it only displaying those effects when dithering is present.

Perhaps you can clarify for me whether CachyOS relies on Linux architecture that has the ability to dither? I’d assume so?

It is strange that those effects are seen on the same wallpapers on multiple OS’s. If it’s not GPU then it’s the E-Ink itself.

I’m also curious as to what the characteristics of those wallpapers are that seems to enable some sort of additional processing.

jordan

markdotpeters5 Yep very frustrating 🙁
I am using cachyos with my rx6600 which is dither free out of the box once you finish the installer/setup. I haven't tried the intel arc on it but I head from a linux intel dev awhole back that it only dithers on 6bit lcds but its spatial not temporal which is good. I think theres a way to disable it though. Nvidia of course also is good once you disable dithering with x server settings app thats built in. (you have to set a config to load your settings each boot for nvidia)

Aquila I was wondering about this so maybe ill try it sometime soon just tired of loading different installs lol.

Aquila

Probably old versions are dither free?

jordan

These windows wallpapers dither 100% - If anyone is up for setting each as their wallpaper in windows im sure you will see exactly what I saw.

jordan

Another update.. When viewing these wallpapers from cachyos in the browser im also noticing similar movement.. so is windows not dithering? im so confused. eink should not refresh on static content so what about these wallpapers is causing this? I am now wondering if its this boox monitor that is somehow being triggered by those wallpapers.. ugh sorry guys. Its weird other photos I view I havent noticed that movement.

jordan

@WhisperingWind Can you test those wallpapers on your capture card?

WhisperingWind

jordan

I checked the images you mentioned. I haven’t observed any temporal dithering on Windows 10 22H2 when the GPU driver’s dithering setting is left at its default. I observed temporal dithering in the HDMI recorder’s capture only when I set the dithering mode to DITH8 or DITH10.

The GPU settings were RGB 4:4:4, 8‑bit.

A bit of hardcore tech 🙂

Temporal dithering is set by the graphics driver and runs in hardware - usually in the GPU’s display path or in the panel’s own controller (TCON). The OS does not paint “temporal” noise into the image. In the pipeline, dithering is the last step before the signal goes out over HDMI/DisplayPort/etc. Before that, things like scaling, gamma/lookup tables, color conversion, and blending are done at higher precision. That’s why dithering belongs at the end, in the GPU.

If you do it earlier (say, in the OS), the GPU receives frames that already flicker on purpose. Temporal dithering changes a pixel slightly each frame so your eyes see a color in between the available ones - even a still image changes every frame.

Those frame-to-frame changes make later color processing on the GPU (tone mapping, gamut mapping, gamma) less reliable, because the GPU no longer sees a steady color per pixel. To undo that, you’d have to de-dither over time, then do the color processing, then dither again. That adds delay, loses quality, and is rarely practical. So OS-level temporal dithering is pointless; the GPU needs a clean, undithered signal.

For an RX 6600, the hardware unit that handles dithering is called FMT (formatter). The Linux amdgpu driver configures it and turns algorithms on/off. Windows works similarly.

P.S. On the plus side, we discovered a new, unexpected behavior in E Ink displays that’s worth paying attention to when choosing such a device. That’s valuable.

JTL

AshX It is strange that those effects are seen on the same wallpapers on multiple OS’s. If it’s not GPU then it’s the E-Ink itself.

I don't want to poo-poo every attempt at research, but I've also wondered if other E-ink processing/artifacts could be mistaken for GPU sided dithering.

jordan

The display mode I was using on the Boox Mira Pro has a bug that triggers the monitor to change its own dithering even when displaying static content, which shouldn't happen. This is why I thought it was an OS issue—eink displays are completely static and retain whatever image is shown until the video output changes (you can even unplug the monitor and it keeps the image).

The spatial dithering modes on eink displays are used to show gray shades. After more testing, one of the preset modes is definitely bugged and causes movement/dithering on certain content, such as the wallpapers I listed. When I switch to video mode (which applies full-screen spatial dithering) or high quality mode (which uses zero dithering), I don't see any movement. In fact, when I enable GPU dithering, I can actually see temporal dithering activate in the non-bugged modes—it goes from completely static to chaotic movement/temporal dithering.

I apologize for the mistake—I had no idea the Mira Pro had a bugged mode. You can still use eink to detect temporal dithering, but you need to avoid modes with this behavior. The high quality mode is probably best for testing since it doesn't use dithering for gray shades. It makes the screen flash with any movement (eink's slowest refresh mode), so it actually works well for detection. My initial findings weren't accurate because that bugged mode threw me off.

AshX In CachyOS with the 1660 Super, I observe zero dithering when I disable dithering in NVIDIA X Server Settings. When I enable dithering, I see constant temporal dithering no matter what content is shown, which is different from what I originally observed due to the bugged display mode.

JTL When the screen is static with eink, there should normally be zero movement, but it's 100% bugged for this display mode since the others don't do that. I can actually activate GPU dithering, and that's when those non-bugged modes start to show dithering when static.

WhisperingWind Thank you for checking those wallpapers! Do you happen to know the command for forcing dithering off for amd in Linux? I don't see any dithering on the RX 6600 with my setup, but I'm unsure if things will be different on other displays that may advertise color depths that require FRC.

So after all of this I can say that the RX 6600 and 1660 Super do not use temporal dithering in Windows 1809-22H2 and Windows 11 25H2 while using a true 8-bit display, since what I was seeing was the same type of dither logic through each one that I found out was all due to that bugged display mode. Now, if you use a different display, I'm not sure how that would change things, as we know 6-bit or FRC normally triggers dithering. I still think eink can still be used as a test tool, but make sure to test each mode and ensure none show movement when things are static. I also tested with a Dasung 253 monitor and didn't see any bugged modes so this is a Boox issue, I'm glad the other modes appear fine on the Boox.

AshX

JTL don't want to poo-poo every attempt at research, but I've also wondered if other E-ink processing/artifacts could be mistaken for GPU sided dithering.

It seems plausible. Especially since we don’t really have a way to measure dithering on E-ink in the same way we do on traditional screens. I still think it’s a useful diagnostic tool, but probably not one you can rely on as your only means of measurement.

« Previous Page Next Page »