• 1
  • 6
  • 7
  • 8(current)
  • 9
  • 10
  • 42
4K HDR10 - State of Play - important media player limitations - LAST UPDATE sept 2020
(2018-08-19, 09:19)wesk05 Wrote:
(2018-08-19, 06:37)Mount81 Wrote: So what would be your opinion and conclusion here regarding which has just dithering and which may have true 10bit?
Did you switched ON that "manual 10 bit" option in OSMC?
Vero 4K and LE-S905X are the good ones. The patterns that you see are due to rounding errors. CE-S905X is actually the worst. It is actually 8-bit with dithering. CE-S912 is 10-bit with dithering. If there wasn't dithering CE-S192 will also look like Vero 4K and LE.

Yes, 10-bit mode was enabled for Vero 4K.
(2018-08-19, 09:07)atomizasser Wrote: Which version of CE was used?
8.95.0  release   
 Well, I have based my judgement on this previous post, and on the pictures also here, as you can read it identifies and discriminates in a rather opposite way between bad and wrong result. I thought that if you get more tiny and more separated blocks, than it's closer to the good result, and if it show less separated and huger blocks, or just rather columns than it's rather worst. Or I just didn't get it right how to conclude the difference?

Edit: taking a closer look, now I see what you mean. S912 and CE-S905X have some "noisy" patterns, with some color fibrillation and fragmentation in them and not so homogenous colored blocks. That may be the sign of the bad dithering? No such mentioned in the Quants2D test's original page, it just differentiates them by the size and, number of the blocks, columns.
Reply
(2018-08-19, 09:19)wesk05 Wrote: Vero 4K and LE-S905X are the good ones.
@wesk05, you might want to try the LibreELEC-S912.arm-8.2.5-Chroma.422.10bit.img.gz image in the following folder:

Mega - LE 8.x testing Folder (click)

Leave the GUI at 1080p and remember to enable the Settings > System > AMLOGIC > 422,10bit option
Refresh / Resolution switching must occur for AML LE to set 422, 10bit automatically before the start of video 4K HDR playback, otherwise you hard Set it via SSH. Wink

Reply
(2018-08-19, 09:38)Mount81 Wrote: I thought that if you get more tiny and more separated blocks, than it's closer to the good result, and if it show less separated and huger blocks, or just rather columns than it's rather worst. Or I just didn't get it right how to conclude the difference?
Edit: taking a closer look, now I see what you mean. S912 and CE-S905X have some "noisy" patterns, with some color fibrillation and fragmentation in them and not so homogenous colored blocks. That may be the sign of the bad dithering? No such mentioned in the Quants2D test's original page, it just differentiates them by the size and, number of the blocks, columns. 
Dithering on CE-S905X is what is making those squares look separated. From a purely visual prespective that is actually fine, but the purpose of my testing was to check true 10-bit output and CE fails in that test. It seems to be a kernel problem because the output is good on both Vero 4K and even LE on the same box (Mecool M8S Pro+).

Since dithering changes with frame and this being a static test pattern, dithering produces the flicker that I mentioned earlier. You are unlikely to notice this flicker in normal viewing (unless there is a static scene).

(2018-08-19, 09:44)wrxtasy Wrote: @wesk05, you might want to try the LibreELEC-S912.arm-8.2.5-Chroma.422.10bit.img.gz image in the following folder:
Leave the GUI at 1080p and remember to enable the Settings > System > AMLOGIC > 422,10bit option
Refresh / Resolution switching must occur for AML LE to set 422, 10bit automatically before the start of video 4K HDR playback, otherwise you hard Set it via SSH. Wink 
I will check it out.
Reply
(2018-08-19, 09:44)wrxtasy Wrote:
(2018-08-19, 09:19)wesk05 Wrote: Vero 4K and LE-S905X are the good ones.
@wesk05, you might want to try the LibreELEC-S912.arm-8.2.5-Chroma.422.10bit.img.gz image in the following folder:

Mega - LE 8.x testing Folder (click)

Leave the GUI at 1080p and remember to enable the Settings > System > AMLOGIC > 422,10bit option
Refresh / Resolution switching must occur for AML LE to set 422, 10bit automatically before the start of video 4K HDR playback, otherwise you hard Set it via SSH. Wink         

Isn't this is 12-bit 4:2:2 output that is being enabled - not 10-bit?  It will be 10-bit content carried in a 12-bit signal, but the signal itself will be 12-bit?  There is no 4:2:2 10-bit 2160p output mode in the HDMI specs AIUI - only 12-bit is supported for 4:2:2 at that resolution?

Reason I make that distinction is that any HDMI device like an AVR or HD Fury Vertex that reports bit depth will presumably report 12-bit (as that's the signal bit-depth) not 10-bit (which is the content bit-depth - with I assume 2LSBs of 0s?)  If people with this set-up see 12-bit or 36-bit reported they may think there is a fault.

wrxtasy - if this will run on a Minix U9-H I can give it a go and see?

EDIT - gave it a punt on my Minix U9-H and it is outputting 4:2:2 12-bit (reported by my HD Fury Vertex) as expected with 2160p23.976 HDR 10-bit content when the 10-bit option is selected in the AMLogic display menu. With that option not selected I get 4:4:4 8-bit output (but with an HDR EOTF flagged and HDR Metadata - as previously)

I guess the options you'd want for HEVC 10-bit HDR replay of 23.976p content are either : 4:4:4 10-bit or 4:2:2 12-bit (as there is no HDMI standard for 4:2:2 10-bit).  However 4:2:2 12-bit is also valid for 2160p50/59.94 content - whereas 10-bit output is only supported for 4:2:0 - so 4:2:2 12-bit output for all frame rates makes a lot of sense.

(https://www.hdmi.org/manufacturer/hdmi_2...q.aspx#146 shows that there is no 4:2:2 10-bit mode for any UHD format)
Reply
(2018-08-20, 18:55)noggin Wrote: Isn't this is 12-bit 4:2:2 output that is being enabled - not 10-bit?  It will be 10-bit content carried in a 12-bit signal, but the signal itself will be 12-bit?  There is no 4:2:2 10-bit 2160p output mode in the HDMI specs AIUI - only 12-bit is supported for 4:2:2 at that resolution?
I believe if the internal processing is at 10-bit depth, manufacturers use 4:2:2 10-bit to indicate that the output is actually 10-bit (with zero padding). If the internal processing is 12-bit, the output can be true 12-bit or zero padded 10-bit.
Reply
(2018-08-20, 21:14)wesk05 Wrote:
(2018-08-20, 18:55)noggin Wrote: Isn't this is 12-bit 4:2:2 output that is being enabled - not 10-bit?  It will be 10-bit content carried in a 12-bit signal, but the signal itself will be 12-bit?  There is no 4:2:2 10-bit 2160p output mode in the HDMI specs AIUI - only 12-bit is supported for 4:2:2 at that resolution?
I believe if the internal processing is at 10-bit depth, manufacturers use 4:2:2 10-bit to indicate that the output is actually 10-bit (with zero padding). If the internal processing is 12-bit, the output can be true 12-bit or zero padded 10-bit. 

Yep - makes sense but potentially confusing when the output format is 12-bit and reported as such.
Reply
(2018-08-20, 18:55)noggin Wrote: Isn't this is 12-bit 4:2:2 output that is being enabled - not 10-bit?
Yes the Kodi option should be labelled 422,12bit - that being the ouput signal bit depth. I will change it in the options settings to prevent user confusion.

Quote:I guess the options you'd want for HEVC 10-bit HDR replay of 23.976p content are either : 4:4:4 10-bit or 4:2:2 12-bit (as there is no HDMI standard for 4:2:2 10-bit).  However 4:2:2 12-bit is also valid for 2160p50/59.94 content - whereas 10-bit output is only supported for 4:2:0 - so 4:2:2 12-bit output for all frame rates makes a lot of sense.
Yes 422,12bit results in a lot better compatibility across a range of Projectors, AVR's & Soundbars. My Yamaha for example works with 422,12bit but I get a black screen when using 444,10bit after refresh/resolution switching and trying to watch 24p content.

PS: You should find Multichannel LPCM audio is now output as 24bit as well, thx. to @graham8 from Vero/OSMC.



@wesk05, I was mainly interested in the Quants2D test patterns picture testing with LE-S912-8.2.5
Should be pretty similar to Vero 4K

Reply
I guess it should auto detect the source and be able to switch to 444 / 10bit for <30Hz and 422 / 12 bit for >30Hz. Or possibly could 422 / 12 bit also with <30Hz make more sense? I've read a couple of times that 12bit output can make the color transition a bit better, even if both the source video and the TV is still just 8 or 10bit (just like in the case of 10bit compared to 8). Or the "closer" Chroma 444 setting would have more positive effect on the overall PQ, than the benefit of 10->12bit "smoothening"?
Reply
(2018-08-21, 08:04)Mount81 Wrote: I guess it should auto detect the source and be able to switch to 444 / 10bit for <30Hz and 422 / 12 bit for >30Hz. Or possibly could 422 / 12 bit also with <30Hz make more sense? I've read a couple of times that 12bit output can make the color transition a bit better, even if both the source video and the TV is still just 8 or 10bit (just like in the case of 10bit compared to 8). Or the "closer" Chroma 444 setting would have more positive effect on the overall PQ, than the benefit of 10->12bit "smoothening"?

Hmm - I guess some displays will do different processing on a UHD signal with 10-bit content presented as 10-bit (at 4:4:4 for <=30Hz or 4:2:0 for >30Hz) vs 10-bit content presented as 12-bit (at 4:2:2 for any frame rate or 4:2:0 for >30Hz)?  

However some displays seem not to like 4:4:4 10-bit content at any UHD frame rate, whilst others only accept 4:2:0 10-bit at higher (50/59.94/60Hz) frame rates.  (Many Sources, AVRs and TVs have options that enable 'Enhanced' modes for UHD HDR capable HDMI inputs/outputs that mean switching to 4:2:2 12-bit from 4:2:0 10-bit I think. By default these are usually disabled - meaning 4:2:0 10-bit is the default choice?)

I'm guessing the safest combination if you can't do full EDID-based output selection may be 4:2:2 12-bit (carrying 10-bit) for <=30Hz and 4:2:0 10-bit for >30Hz as these are the lowest bandwidth outputs available that won't truncate bit depth?  I may - of course - be very wrong.
Reply
This compatibility might be the reason that some media players (like my Q10 Pro), that can't do 12 bit output switch rather to 422 / 10 bit and not to 444 / 10 bit on 4K <30Hz?

Some also experienced, that if 4K / 420/10bit/HDR is the setting for the GUI, than it will switch in the case of i.e. 23.976fps videos to 444 and HDR, but the TV's or AVR'S info won't show anything regarding the current input bit depth, neither 8 or 10bit. What this mysterious output would supposed to be? In what case would it happen, that no bit depth information is send/shown? And what would be the reason to apply it?
Reply
Forgive me if this is a stupid question or if it's in the wrong thread, I'm kinda new with Kodi. 
On my Minix U9-H (both in Kodi 17.6 with Android and LibreELEC) 4K 10bit HDR material is output as YCbCr 4:4:4 (according to my Pioneer receiver) even if it's 4:2:0. Is that oke? In android I've set the display to 10bit 4:2:0.
Reply
(2018-08-25, 10:30)Huit Wrote: Forgive me if this is a stupid question or if it's in the wrong thread, I'm kinda new with Kodi. 
On my Minix U9-H (both in Kodi 17.6 with Android and LibreELEC) 4K 10bit HDR material is output as YCbCr 4:4:4 (according to my Pioneer receiver) even if it's 4:2:0. Is that oke? In android I've set the display to 10bit 4:2:0.
 That's absolutely fine. Moreover the 444 output is the best in any case. The 420 has to be upsampled to 444 anyway (and than to RGB, but that's the business of your display) to be able to be displayed by your TV. So the upsampling directly to 444 is always better than when it's just being upsampled to 422 as it not involves any unnecessary intermediary step.
Reply
Thanks! That was a clear answer.
Does this mean I also need to activate the Chroma subsampling 4:4:4 on the input of my LG C8 OLED? (Rtings mentions how to do this here, scroll down to "Miscellaneous settings" https://www.rtings.com/tv/reviews/lg/c8/settings )
Reply
(2018-05-25, 07:50)wrxtasy Wrote: Xiaomi Mi Box (Oreo OS) running Kodi Krypton (click):
  • WARNING ! - the Mi Box Oreo Firmware has quite a few serious Bugs, see the Mini Review HERE (click)
  • No Gigabit Ethernet for high bitrate 4K video playback - a cheap USB3 > Gigabit adapter needs to be used to make high bitrate streaming bulletproof. ASIX (Ugreen) & Realtek chipset (USB Hub) adapters confirmed working.
  • No auto refresh or auto resolution switching. 2160p60 or 2160p24Hz must be manually selected in Android Settings for 4K
  • Multiple options in Android Settings on the Oreo Mi Box for auto HDR to SDR or auto SDR to HDR switching as well as Chroma & bit depth selections.
  • 4K HDR > SDR produces decent picture outputs.
  • Kodi Krypton & Leia PVR Addons produces poor quality (half motion) deinterlacing for OTA TV viewing.
  • 60Hz 4K HDR YouTube works if the community modded Webwrapper Smart Youtube TV App (click) is installed.
  • Amazon (Prime) 4K HDR Video works if the v.4.7.9 Prime Video App (click)[/url] is installed.
 New to the forum... this doesn't seem to be working for me. I have changed the settings for HDR to SDR to "on" in the Mi Box itself but the colors are still way off. Are there specific settings in Kodi that'll help with conversion? Would this require Kodi to cause the server to transcode? I'm running Emby server.
Reply
Sad 
(2018-08-19, 20:04)wesk05 Wrote: Dithering on CE-S905X is what is making those squares look separated. From a purely visual prespective that is actually fine, but the purpose of my testing was to check true 10-bit output and CE fails in that test. It seems to be a kernel problem because the output is good on both Vero 4K and even LE on the same box (Mecool M8S Pro+).

Since dithering changes with frame and this being a static test pattern, dithering produces the flicker that I mentioned earlier. You are unlikely to notice this flicker in normal viewing (unless there is a static scene).
Hi @wesk05 I read about this 8 bit to 10 bit "dithering" on s905x a long time ago and because of that I bought a s912 box instead of a Vero 4k.

I tried CoreELEC, LibreELEC and a few others but could never get a good HDR image.

I left it unplugged for months and have only recently started reading about it again, it seems that the problem I was having is every time i tested HDR i used Terminator 2 and the picture was REALLY dim.

I found out yesterday that this and a few other titles are missing luminance data and OSMC has recently fixed this by changing fallback luminace from 5000 to 1000 nits.

I spoke to CoreELEC about this and they are adding the fix to their build also.

So back to my original thought.

When I read about this 8bit to 10bit dithering on s905x it said that it was a hardware limitation.

Reading through this thread last night you seem to be saying that this is not the case and it is a software limitation ?

Also you have said the CoreELEC builds for s905x and the s912 both suffer from this ?

Basically I am REALLY REALLY confused and would like some clarification from someone who knows.
  1. Do any of the s905x boxes have a hardware limitation preventing them from passing through true 10 bit video ?
  2. What about s905d and s905w?
  3. If the s912 does not have this limitation why does the s912 CoreELEC build have a problem ?
  4. The original vero 4k has a s905x right ? Does the vero 4k+ have a s905d ?
  5. do the latest builds of libreelec based on Kodi Leia only support s905x and not s912 ? will they ever support s912? and how about s905d and s905w ?
I mentioned the stuff about terminiator 2 and the dim picture because I also thought it didn't look anywhere near as clear or sharp on coreelec as my tvs internal player.

Could this be dithering ?

What is dithering exactly ?

are 8bit to 10bit even related to dithering ?

I am again tempted to buy a Vero 4k+ but this whole 8bit to 10bit conversion thing has me so confused and is putting me off.

Plus I have a s912 and a s905w sitting around doing nothing.

I wish my tv had earc and could just pass true hd back though the hdmi Confused

Sorry for the long post.
Reply
  • 1
  • 6
  • 7
  • 8(current)
  • 9
  • 10
  • 42

Logout Mark Read Team Forum Stats Members Help
4K HDR10 - State of Play - important media player limitations - LAST UPDATE sept 20209