Feedback Color issue thread - Discussion 2

Discussion in 'ZIDOO X9S' started by n_p, Jun 30, 2017.

  1. wesk05

    wesk05 Member Beta test group

    They did contact me asking whether I would be willing to test the firmware that is supposed to fix this issue. I replied yes and after that I never heard back anything.

    I will see whether I can test this new firmware tonight.
     
    boblo likes this.
  2. wesk05

    wesk05 Member Beta test group

    Desktop color issue (RGB Full to YCbCr conversion) is definitely fixed in this experimental firmware (1.48). I checked primary, secondary colors, color checker, grayscale, YouTube & Playstore icons. They are all correct now. I cannot reproduce any of the issues that @n_p has reported. The output is 16-235 for both YCbCr and RGB modes. There is no passthrough of superwhites (236-255) or superblacks (0-15) now.

    There is also another change. RGB quantization bits (Q0,Q1) are now set to limited range. This is first time I am seeing those bits being used in an Android device. This doesn't necessarily do any harm. The one issue setting the RGB output to limited is folks like @boblo won't be able to use the device as a mini pc. Also, RGB color space doesn't seem to apply to 4K 50/60Hz. I can't remember now whether this was the case before.
     
    DolbyVision likes this.
  3. spring

    spring Guest

    Hi wesk05
    RGB and YUV also have the limited mode and full mode, we output the limiter mode or full mode depend on the EDID of display.
    if the EDID support full mode, we will output the full mode, and if it only support the limited mode, we will output the limited mode.
    this is the HDMI EIA-CEA-861 spec.
    edid.jpg
     
  4. spring

    spring Guest

    do you set the deep color of 4k 50/60?
    Unbenannt.png
     
  5. n_p

    n_p Active Member

    That explains why YCbCr modes in my case only output full range - which is highly unconventional. I would make this a manual setting if possible and not make it EDID dependent.

    Here is the logic for it - TV capability means nothing, if for example you have all your inputs (or some of them) calibrated for a limited signal level (which again is actually the AV standard). Most calibrators for example would calibrate for a limited signal standard - as youve set it up right now - I have no ability to put the box in a limited output mode - manually.

    Thats forcing full range output over limited range output - just because of capability. Which is highly unconventional.

    That said, if what spring said is the case - then it explains, why in my case - all YCbCr output modes output a full range signal.

    Also as you can see above - the same "overreliance on EDID" in wesk05s case now means that he has issues getting full range into his equipment - so overall, it might be more sensible to make full range a manual toggle for YCbCr 444 and 422.

    Also - to emphasise that point as well - when manually setting the X9S to YCbCr 444 or 422, or RGB, - although I would always get a "full range signal" - the primary color placement was correct. (But it was so before your "fix" (for a full range signal) - so I didnt put the emphasis on it).
    -

    That explains one half of my issue with the test firmware - the other one is still unexplained.

    When set to auto (color space), the X9S would output a limited signal to my 2016 LG OLED - although it is "full range capable". (TV set at the time was set to limited range mode.)
    When set to auto, the X9S would output a bt2020 color space (?)to my 2016 LG OLED - although the X9S was set to 1080p60. There is NO color standard that would output bt2020 @1080p, so this is wrong behavior. The X9S SHOULD NOT output a bt2020 signal (even if the EDID says that the TV is capable of displaying it), when its output is set to 1080p (or below).

    This happend both with the LGs 4K deep color modes enabled and disabled for this particular HDMI input.

    Also - looking at the bt2020 triangle - magenta is way off - but lets take this one step at a time...
    -

    Over reliance on EDID is probably not a good thing. If you decide to go with that as a default, at least make it overridable via a manual toggle. Please.

    Also I'd advice against making it (signal range) an EDID based default - and here is why. Default for AV equipment is limited (16-235). Which means that the default setting all TVs is either AUTO, or limited. On my LG OLED - when using a X9S there is no auto option (again - handshake...), so the default is set to limited. So for most users it would be set incorrectly out of the box, if their AUTO option doesnt work. This is something you have to decide on your own, there are pros and cons for each approach.
     
    Last edited: Jul 21, 2017
    DolbyVision likes this.
  6. n_p

    n_p Active Member

    Also - I think I have to apologize for outright trashing the changes you have made in this test firmware.

    The results I've got (signal output was always full range) indicated to me, that Zidoo scrapped the YCbCr signal modes, and just set everything up to default to Full RGB, then called it a day.

    Actually their thought process was a bit more refined - but it also lead to those results (only ever getting a full range signal, using either manual color mode setting.).

    And in the only limted range output I could get from the box (using auto) the color space was set to bt2020 (probably EDID related), even though the output resolution was set to 1080p which is something thats not supposed to happen either.
     
  7. n_p

    n_p Active Member

    A few more things that are noteworthy about the beta firmware.

    When I connect the X9S to my 3D Lut Box - it now outputs a limited signal (16-235(?)) across YCbCr 444, 422 and RGB.

    (I had an issue before, where the Lut Box would warp input colors and this is still present, for both YCbCr input modes - but it is gone with RGB limited (YAY - because this would be my main use case for the X9S) -- so please whatever you do - keep the RGB limited mode as an option. :) )

    When using RGB 444 limited, the black level between the GUI interface and the video in Kodi/ZDMC is misaligned. Not by much (ie its not a 0 or 16 issue) but by quite a few ticks. Even more problematic, the Video Level of ZDMCs own player is different than the black level of ZDMCs Kodi Player (DVD Player) by quite a few ticks of the brightness slider.

    GUI Black level in RGB 444 limited on the other hand is now EXACTLY where it should be, for a limited source (based on "typical" input settings for that HDMI port).
    -

    Also as encouragement - here are the CCSG measurements for RGB 444 limited (running through my 3D Lut box, which negotiates "limited" - but without a 3D Lut applied), for GUI colors:

    [​IMG]

    They are pretty much exactly where they should be, based on the performance of the TV.
    -

    To simplify -

    - Please keep RGB 444 limited as an option.
    - Consider implementing toggles for Full/Limited that would override EDID defaults.
    - Look into the issue where Color mode Auto would result in a bt2020 color space, even when 1080p60 (or below) was selected as output resolution.
    - Investigate, why in this case (auto) signal level was limited, even though the connected TV was 0-255 compatible (and all other X9S display modes "saw" it)
    - Look into the issue of "slightly" (but enough to be obvious) misaligned black levels between GUI, ZDMC (default player) and ZDMC (kodi player). Noticed with RGB 444 limited. Other color modes not tested.
     
    Last edited: Jul 21, 2017
  8. mrs4t4n

    mrs4t4n New Member

    So, i have an Lg Oled 4k (E6V), the x9s connected through denon x1300 avr.
    It's safe to update to the new experimental firmware?
    Which will be the *best* settings for my setup?

    Thanks in advance!
     
  9. HaoSs

    HaoSs Well-Known Member

    Wait for proper firmware release,
     
    mrs4t4n likes this.
  10. wesk05

    wesk05 Member Beta test group

    I know very well what is in CTA & HDMI standards. You can see that I talk about Q0,Q1 bits in my post.

    If indeed the quantization range is being set based on sink support and source content, I will have to recheck with a test pattern set to full range. There may be random clips out there, but not a single commercial video will have the range set to full. Almost all TVs produced in the last several years will have QS=1 in their EDID. So with a full range source, you are expected to get full range RGB output from X9S. QY=1 can be seen only on Sony TVs and a few other brands.

    4K 50/60Hz RGB doesn't support Deep Color. If you are referring to whether it has been enabled for the HDMI port on the TV, then yes. nVIDIA Shield connected to the same HDMI port has no problem switching between 4K 50/60Hz YCbCr 4:2:0 and RGB (8-bit).
     
    Last edited: Jul 21, 2017
  11. wesk05

    wesk05 Member Beta test group

    Can you check this pattern and see whether you can actually see the flashing bars below 16? If you can't see it with your normal setting, max out brightness (or black level on newer Sony TV) and try again. If the output is indeed full range, you should be able to see bars below 16. If the output is limited, you will not see the bars even with brightness (or black level) maxed out.

    AVS HD 709 Black Clipping Test Pattern

    It is OK for YCbCr output to have full range. What will throw displays off is, the quantization range bits (YQ) set to full.
     
  12. n_p

    n_p Active Member

    I see black bars below 16 flashing for both

    - directly connected to the LG B6 OLED (black level setting has to be set to "high" on the TV to get "close to bar 16 = black". High is the setting the TV has to be set to when you feed it a FULL (0-255) signal - so I believe I get a 0-255 signal at that point.

    - connected to my 3D Lut Box (black level setting has to be set to "low" on the TV to get to "close to bar 16 = black") - so I believed, that this was a limited signal reaching my TV - that said, the bars below 16 are still flashing, and are NOT clipped.

    [EDIT: I just remembered, that there is the black level mismatch between GUI and the Kodi player. Video black level in Kodi is "too bright" with the X9S outputing a limited signal (it is correct for GUI black). This could be why I see below 16 flashing even when I believe the X9S is outputing a limited signal.]

    So what I believed to be a correct limited signal probably is NOT. (?) (Because in both cases bars lower than 16 are flashing.)

    (The 3D Lut box is capable of both Full and Limited signal handling, but I dont know what its EDID communicates.)

    Selecting YCbCr 444, 422 or RGB 444 doesnt make any difference.
    -

    What you would do conventionally is to have

    YCbCr 422 and 444 fixed at limited by default (real limited so with clipping) then have a toggle for superwhite - yes/no that doesnt change the black level (still limited (16)).

    RGB 444 fixed at full by default (0-255), but then have a toggle that allows you to set it to "limited" (or whatever Zidoo currently does, when I connect the X9S to my 3D Lut box (black level shifts by a full>limted amount, but below 16 still vissible (edit: in video))) - because thats the only setting I dont get the color shift introduced by my 3D Lut Box. ;)
    -

    That would be how you design those settings conventionally and without EDID lookup.
    -

    The 3D LUT Box is a bad way to determine "limited signal behavior"; but its the only way I currently have to "switch it (limited/full) for test purposes" - as the box is even able to do limited to full conversions (I dont have this matrix loaded on the box currently... ;) ), so what I am saying is - that I dont know how "prestine" the output signal is, compared to the input signal, at the point where it reaches my TV.

    All I know is, that for the 3D Lut Box input level is supposed to be output level (input level == output level), the way I use it currently.

    Having the EDID limited/full decision not "overideable" with an OS toggle is just asking for problems in my opinion. I know no manufacter whatsoever, that doesnt allow for any manual way to switch between limited and full (if the device allows both for limited and full signal output). This is asking for disaster. Imho.
     
    Last edited: Jul 22, 2017
  13. n_p

    n_p Active Member

    Here is another description that doesnt rely on an a VIDEO testfile. :)

    Used the black level test on http://www.lagom.nl/lcd-test/ with the built in Chrome browser.

    HDMI black level calibrated according to other source devices.

    - connected directly to the LG B6 TV - TV black level set to high (= full range setting):
    steps 1-5 are invisible (become visible when I crank the TVs brighness setting), 6 is barely visible, 7+ are clearly visible.

    - connected through my 3D Lut Box - TV black level set to low (= limited range setting):
    steps 1-5 are invisible (become visible when I crank the TVs brighness setting), 6 is barely visible, 7+ are clearly visible.

    So, connecting the X9S to my 3D Lut box shifts the black level exactly by the full>limited amount. Yet below 16 values are still visible. (And I dont necessarily know why).

    Do you see the issue with relying on EDID exclusively? You cant see what it is set to. You are only guessing. You have to have an override in the form of a toggle for limited/full. That then forces limited andor full.
     
  14. wesk05

    wesk05 Member Beta test group

    I rechecked this with EDIDs that I know have RGB/YCC quantization declared. As @spring mentioned earlier, in this firmware the output range is indeed set based on the sink's EDID. If the sink declares QS/QY=1, the output is set to RGB Full or YCbCr (Full). If the sink declares QS=1, but QY=0, then the output is RGB Full, YCbCr (Limited with clipping). If the sink doesn't declare any quantization (QS/QY not set), then the output is RGB and YCbCr Limited with clipping.

    The table below shows the results for the different configurations.

    [​IMG]
    * - 16-235 is scaled to 0-255. Superblack (0-15) and superwhite (236-255) is clipped.
    ^- 0-255 is scaled to 16-235. 0 is mapped to 16, 255 is mapped to 235.
    # - unusual scaling, superblack (0-15) is clipped, 221 and above is also clipped. This in effect reduces the range to 16-221.

    YCbCr (YUV) input means the source is a test pattern video with superblack/superwhite, but VUI set to limited range. The test pattern was played in ZDMC. VUI when set to full didn't change the results. So, it looks like that information is not being taken into consideration when setting the output.

    For RGB Limited/Full the input is Spectracal MobileForge app.

    Overall, it is fine except when the sink declares selectable YCC quantization (QY=1). HDMI standard does define setting the quantization range of the output based on what the sink declares, but I think an override needs to be provided. To my disbelief, I found that the newer Samsung SUHD TVs actually don't declare RGB quantization in its EDID. So, if you have a Samsung SUHD TV, X9S will always set the output to RGB Limited even though the TV actually supports RGB Full.
     
  15. wesk05

    wesk05 Member Beta test group

    This is interesting. I cannot see any bars below 16 flash. I have tried it on Samsung/Sony TVs. Sony TV has selectable RGB/YCC quantization and so X9S output is full for both RGB and YCbCr. Analyzer readings show that superblack (0-15) is clipped in all cases. Analyzers had the EDIDs from Samsung/Sony TVs and a custom "Mega EDID" with every possible option allowed in CTA 861-F/G, HDMI 2.0a/b loaded.

    What exactly do you mean by black level mismatch between GUI and Kodi player?
     
  16. n_p

    n_p Active Member

    In ZDMC (Kodi as well, but lets use ZDMC as the example here, because of obvious reasons.. :) ) -

    - in the ZDMC built in player and in
    - Kodis default player (wihin ZDMC)

    for me black is displayed too bright (I have to shift brightness down by 5 (and anothr amount on the other player) "ticks" on the TV, for the 16 bar to == black. This is also visible in movies (black bars are "glowing" - this is very easy to diagnose on the OLED, because they should be perfect black.).
    GUI black level on the other hand is "spot on".

    - this occurs on

    RGB limited and (afair) all the other YCbCr limited options -

    The difference is much less than a full 0-16 "shift", but its notable.

    (So currently I have to shift brightness down, once I start a movie, and back up, once I use the GUI for a longer period of time. The levels (actually all three ZDMC player/Kodi player/GUI) arent matched.)

    And I dont think the 3D Lut Box is responsible for shifting the black level, yet - I can not be sure, because - Zidoo refused to release a firmware where you have any controls over any of the output signals, and then said "test". :/

    I would love to test a limited output signal directly to my TV - but I cant.

    As a point of clarification, did you take the bit level measurements on the GUI level, or from within a "video file" (playing in Kodi/ZDMC)?
     
    Last edited: Jul 23, 2017
  17. wesk05

    wesk05 Member Beta test group

    @n_p : Sorry, I still don't understand what you mean by GUI level. As for measurements, for the YCbCr/YUV category, the measurements are from within the video when playing in ZDMC (external player). I will get to this point shortly. For the RGB categories, the measurements are from the test patch displayed by MobileForge.

    Grr... :mad:. I just found out that ZDMC (external player) and Kodi (Mediacodec) rendering are different. I can see 0-15 bars flashing when played in Kodi 17.3 or with external player turned off in ZDMC. Looks like I will have to spend another few hours testing the patterns in Kodi.
     
  18. n_p

    n_p Active Member

    There we go.. :) Finally matching results. :)

    (Gui Level (OS Level) == "not measured on a video" (but f.e. using Mobileforge instead) - hypothesis here is, that HW accelerated video (?) doest use the same correction table, that Zidoo patched in the test firmware. Yet the goal is to get color output and blacklevel correct (thats usually the case), and blacklevel "synced". :)

    I for one now know why Amazon is forcing limted. Less complexity. ;)
     
  19. Gwadalolo

    Gwadalolo New Member

    I don't allow you to tell me I'm wrong just because you didn't understand what I wrote.
    I have a crappy TV and an excellent Zidoo box for saving HDMI IN and superb NAS features with luci, and I don't regret a euro because other boxes usually don't have such features. I couldn't care less about your color drama, and this is just MY opinion.

    About fixes, all fixes are welcome once done, but of course devs have to define priorities because you can't fix everything.
    Again, please read : I don't tell anyone not to fix this, just that I'd have other priorities.

    As usual, the guys that shout louder and make dramas obtain what they shout for...
     
    nikos_a and Sarco like this.
  20. n_p

    n_p Active Member

    So what you are saying is, that you dont like facts, you are afraid of buyers remorse, but you love feelings and religion, and if someone says anything to pop one of the bubbles you are in ("the excellent Zidoo box - you are using HDMI IN with"), you act on your mood and try to fight factual issues on a pseudo religious level, just so you dont have to align them with your version of reality?

    Cool.

    Also - thanks for telling us how crappy your TV is, and how you use the Zidoos recording feature - because -- those are all nonsequitors, aimed to dilute the issue at hand.

    Also, if people tell you that you are wrong you very much like to tell them, that they just dont understand the love you have for a piece of hardware you bought.

    I ow you an explanation though. I didn't have to react that way. But I am done with people doubling down on ignorance and "fact shaming", just to protect an uninformed buying decision they dont want to come to terms with. When I say that with the bug thats just been addressed, and other bugs concerning color output that Zidoo hasnt adressed so far, the X9S, although marketed as "premium" was outputing signal quality that was unacceptable and worse than all its competitors -

    I dont do this to attack you personally (regardless of how you "feel" about it) I do this to inform potential customers to make informed buying decisions. Thats how "markets" work. (With the exception, that you have to do this in the manufacturers forums, as a buyer, because currently there is no independent platform (journalism) that would actually act in this role, which is not optimal.)

    Technology is not a lifestyle product (regardless of what Apples marketing tells you), if there is an issue, you can objectively prove it, show it, fix it. When there is an issue, you can quantify it.

    "Yes, but I use mine on a crappy TV, and use an entirely unrelated function, which "still is great"." Is no defense for something thats broken. Its not even an excuse.

    And I dont like people getting away with trying to pull a "but what I feel about a product should be more important" to downplay technical issues. Doubling down on that wasnt the best idea either.
     
    Last edited: Jul 30, 2017

Share This Page