Products to try or avoid? PWM Flicker and Temporal Dithering Testing

jordan

Yes, I have it set to discrete in BIOS and under device manager, I only see the Nvidia gpu not the intel iGPU. I am wondering if I am just video recording pixel inversion given mention of this post over at the Stillcolor discussion:

https://ledstrain.org/d/1075-apple-miniled-products-2021/8

This quote from Seagull?

This brings me onto something I think about a lot. Why do people hate intel graphics so much? I think it’s not because intel graphics dither (they don’t, I’ve tested lots of them). I think people get pain from intel graphics because they are not temporally dithering! So they are seeing a strong half framerate flicker, undiluted by temporal dithering. Something I think will be worse on laptops where their high DPI screens will have more voltage irregularities creating a stronger LCD inversion flicker.

Personally, I have a monitor that I can only use with GPUs which dither. If I hook up my non dithering intel GPU I get pain. The above is the best explanation I have for that.

Maxx

I use this: https://carson.com/product/mp-250-led-lighted-pocket-microscope/

I use 240 fps slow motion camera mode on my samsung s10+ but some newer phones can do higher fps which is good for testing the higher refresh rate screens.

You will be able to see pixel flicker from either pixel inversion and dithering or both. The problem is I have yet to find anything where the image is 100% stable.

I also would record at 100% backlight brightness to help factor out PWM and low video quality due to low light noise.

jordan

I have 1809 running on the second hard drive of the 7i and install the latest studio driver from nvidia to enable 240 fps refresh rate (otherwise the panel is stuck at 64Hz). Thanks for the suggestion!
Laptop is set to discrete GPU, color control dithering disabled.

Difficult to say if this is any better at first glance both holistically and under the scope. I'm not able to get the full raw 240 fps video from the phone camera app just a slowed down version at 30 fps on playback. Still, I was expecting to see more flickering (filmed on the middle part of the lagrom gradient). If someone is watching the video, notice how when I move the camera so that the top left flickering pixels enter the center area, the pixels stops flashing? That is because their is less light captured there and the noise of the phone camera sensor is more visible which itself is a temporally flickering effect.

https://www.dropbox.com/scl/fi/09hib3rhiunqq505kfr5d/7i-gradient-windows-1809.mp4?rlkey=kbchwo7dpvyqvbryqx6f5y0p9&dl=0

jordan

Interesting video. I'm now using the 1809 with the manual update installed. The build version is 17763.1098. Is this correct? All I can say is the somehow the maximum brightness is lower than on Win 11 making it seem calmer to me. Through the scope, it is see hard to tell if all I see is just noise flickering from the camera sensor or dithering or inversion (this might be a good sign). If you have testing 1809 to work on your other machines, then this is probably the way to go if/when you decide to test your 7i. Also, in appearance and performance, I turned everything off (adjust for best performance). On the lagrom gradient banding page, I see more banding.

jordan

Thank you!

I installed 1809 with the update on the T480s (with B140QAN2.3 panel) and it has clearly visible banding. The Lagrom gradient even seems to animate the banding and then stop (minimizing and opening the browser would do that). However, under the microscope measuring the gradient, it is almost completely "stable" and clearly more stable than the 7i on 1809 (might be due to low max backlight brightness causing noise). Perhaps now I am seeing mostly noise + pixel inversion.

https://www.dropbox.com/scl/fi/5h2uk8lz26c7u61zsjtqc/t480s-1809-gradient.mp4?rlkey=dm8cyynb18yfamqo9zw6sg730&dl=0

This is just a 60 Hz panel so this is even more interesting. The gpu is the outdated Intel UHD Graphics 620 not Iris Xe. Also the panel can get significantly brighter than the 7i running 1809 making recording have less camera sensor noise.

For the less flicker sensitive who want to check out videos (T480s 60Hz 1809 best case versus Legion 7i 60Hz Windows 1 23H2 worst case videos of a darker transition part of the lagrom gradient):
https://www.dropbox.com/scl/fo/adv3uvxm0tvi6hndxli2p/h?rlkey=kcsyjrq6p5bf6qprdco9gx4ed&dl=0

Tentatively, the 60Hz T480s running 1809 is similar to the 240Hz 7i running1809 and better than the 7i at 240Hz running 23H2. The worst flickering is with the 7i at 60Hz running the latest 23H2 update.

jordan

Yes, the T480s running 1809 and not win 11 is similar to 1809 240Hz 7i. The T480s has supposedly crap AUO B140QAN02.3 panel I stated earlier, not the LG panel listed in the table which I don't have anymore to swap out and test.

My good news for the day: On the 7i using 1809 with BIOS set to dynamic and using intel graphics but with no intel driver installed using MS Basic Adapter at 64Hz refresh rate, I am seeing almost none of the typical slow alternating flicker which I thought was pixel inversion. The image under the scope is stable similar to using 1809 at 240 Hz but using discrete gpu. I cannot set screen refresh rate any higher than 64Hz with the basic MS driver. When I go back to 23H2 on the 7i and still using 60Hz panel refresh, I again see the slow alternating pattern of flicker even when I force 8 bit in discrete gpu in nvidia control panel or leave the bios set to dynamic and using iGPU.

This lead me to believe there is no extra dithering going on the software and gpu side and the output to the panel is 8 bit. With a 8 bit + FRC fake 10 bit panel as you found, this could be another safer alternative. Thanks again for suggesting 1809. Now I have to be careful not to break this setup.

jordan

For the 7i:

If you want to use 60Hz screen refresh, use 1809 with intel gpu only. Even under 1809 with latest Nvidia studio driver with 60Hz screen flickers a lot (this might be fixable with modified drivers).

If you want to use 240Hz screen refresh, 1809 with latest Nvidia studio driver still seems similar to 23H2 with the same Nvidia studio driver (high screen refresh masking my ability to measure dither correctly). So you may not gain much with 1809. I will take a break from this testing lol and try those older Nvidia drivers later.

Using the 4080, I cannot tell the difference between 10 bit versus 8 bit on 23H2 or 1809. I think the latest Nvidia driver forces some kind of dithering anyway. Update: With 23H2 at 240Hz, it is also difficult to tell the difference between discrete mode or dynamic mode. Later, I would like to see if I could get 240Hz working in 1809 using just Intel iGPU.