[TESTING][S905(X)] 10bit/HDR/Dithering Test Builds & Discussion

  • New build uploaded on github releases page with some changes from the amlogic 20170515 release merged in. A side effect of the updates show an improved HDR->SDR conversion, and I included a bit of info on how to customize the brightness grading for it. Removed build due to broken non-passthrough audio.

    I don't see banding, but I do see some form of slim vertical stripes which I think are dithering artefacts. I had to walk up to the TV to see those, though :)

    I'm not sure what this is from, and I get the same thing on Android also on 10bit output with dithering disabled if I focus on some areas in that video. However, some of my actual test videos don't show it very well.

    That's probably it. My TV's internal player does not exhibit banding though. Is dithering supposed to be turned on?

    With it on, we're cutting off two bits of color depth and adding noise (which helps with native banding). Your display could have processing that minimizes it, or we're running into a hardware limitation, or it's something else entirely. We should be able to add and customize a bit of noise through am_vecm, but it's not something I've spent much time looking

    PS: I changed the default kernel branches. The branch these builds are based on is "amlogic-3.14.y-nougat." The "nougat-exp" branch is very experimental/wip, and full of bugs (currently plays video far too quickly due to PTS changes). WIP and testing stuff will start out in this branch, then move to the main branch.

    Edited 2 times, last by johngalt (June 9, 2017 at 8:23 PM).

  • I don't seem to be getting any audio on the new "8.0.2-testing2" build. I'm just using a regular old HDMI connection straight into my TV (no amp or anything). Testing1 was fine.

  • I don't seem to be getting any audio on the new "8.0.2-testing2" build. I'm just using a regular old HDMI connection straight into my TV (no amp or anything). Testing1 was fine.

    Thank you, I definitely broke it in the merge (only tested passthrough audio). Removing the build until I get a chance to fix it.

  • Ok. I've found the culprit.

    My TV has a "Smooth gradation" setting, which i had on low. Funny enough, turning it off (or to high) gets rid of the banding.

    So I'm sorry to have caused confusion here. I consider this build to be just about perfect now. :)

  • I just fixed standard audio again and reuploaded testing2. I messed up and only updated the tarball, then wiped the builds clean for a personal agile build. So there will only be a tar up for testing2, and if you need the image, you'll need to wait for testing3, or upgrade from a previous install to testing2.

    Edited once, last by johngalt (June 10, 2017 at 7:36 AM).

  • No 4k 10 bit video output also with latest release, Error on screen: no video input.

    I tried three hdmi clables

    8 bit 4k video output is ok

    Edited 4 times, last by saltanar (June 10, 2017 at 10:44 AM).

  • No 4k 10 bit video output also with latest release, Error on screen: no video input.

    I tried three hdmi clables

    8 bit 4k video output is ok

    My AVR reports;

    -YCbCr420 24bit for 4K >= 50hz. (8 bit)

    -YCbCr444 30bit for anything else. (10 bit)

  • My AVR reports;

    -YCbCr420 24bit for 4K >= 50hz. (8 bit)

    -YCbCr444 30bit for anything else. (10 bit)

    Thank you, I believe I found the issue from the merge again. I removed the build until I can fix it and verify. Unfortunately I didn't test >=50hz. For now I recommend testing1.

  • Thank you, I believe I found the issue from the merge again. I removed the build until I can fix it and verify. Unfortunately I didn't test >=50hz. For now I recommend testing1.

    ps. Not sure if my TV supports 44410bit 50/60hz at 4K, but i know it supports 422 10 & 12 bits at 50/60hz.

  • ps. Not sure if my TV supports 44410bit 50/60hz at 4K, but i know it supports 422 10 & 12 bits at 50/60hz.

    To double check this isn't a different issue, could you run the following and post the output here? It doesn't matter which release you run the command on (so it's fine if you're back on testing1) cat /sys/class/amhdmitx/amhdmitx0/dc_cap

  • To double check this isn't a different issue, could you run the following and post the output here? It doesn't matter which release you run the command on (so it's fine if you're back on testing1) cat /sys/class/amhdmitx/amhdmitx0/dc_cap

    NexBoxTK:~ # cat /sys/class/amhdmitx/amhdmitx0/dc_cap

    420,10bit

    420,8bit

    444,10bit

    444,8bit

    422,10bit

    422,8bit

    rgb,10bit

    rgb,8bit

  • Thank you, it looks good. I've decided to hold off on future builds until I get the full nougat amports and 20170515 merges working correctly (nougat-exp kernel branch), which means there's a bit of work ahead. For now, testing1 seems to be in a very good state.

  • Thank you, it looks good. I've decided to hold off on future builds until I get the full nougat amports and 20170515 merges working correctly (nougat-exp kernel branch), which means there's a bit of work ahead. For now, testing1 seems to be in a very good state.

    Agree, testing1 works like a charm for me, and I'm very pleased with all your work. A BIG THANKS to you and kazaq, and every one else who contributed to this build.

  • hello, how can i roll back without tar file for the update folder, must i burn sdcard new and do fresh install?

    Your work is great! Thx!

    You can place the .img.gz in the update folder just like a tarball :)

    Edited once, last by johngalt (June 11, 2017 at 9:29 AM).