[jmarshall]

I'll need information about what the limited on/off issue is. i.e. there is likely an interaction (limited will interact with the various decoders).

If it's just "looks grey" that's to be expected if you don't need it. If it looks different when ON between the different decoders, then that's useful information.

If there is no difference between the decoders when it's OFF then that's useful information.

[Piers]

With limited disabled all looks good.

Enabled using software it looks good.
Enabled using DXVA related and it looks bad.

The the problem appears to be DXVA - your previous build resolved the issue (as far as I can tell) mentioned at the start of this post and in the post I linked to.

I'm surprised more people haven't noticed it.

[jmarshall]

"Looks good" versus "Looks bad" isn't what I'm after

I'm after whether the effect is the same for the UI and the video, or whether the video looks as though the limited colour stuff is applied twice or something.

(It shouldn't be, but who knows - I'm coding blind).

You can probably tell this by having a black-ish skin in use and then start up a video such as Homeland with a black backdrop and see if the black level changes when you do so.

Also, I only touched DXVAHD - I didn't touch DXVA which may or may not be also broken, so there's two separate things to test there.

Cheers,
Jonathan

[Piers]

Oh sorry, I mean't that it does appear the limited range is applied twice, resulting in distorted brightness and contrast. Thi doesn't appear to apply to the UI - only the video playing.

[awskip]

I can also report that the build you posted fixes the issue. No combination of acceleration setting results in elevated black levels. Before either setting DXVAHD on or Allow DXVA2 would show the issue.

UI levels have always been okay on my end (and continue to be) so can't comment on that. Issue was only while playing video.

[_Shorty]

Well, today I spent some time playing with the XBMCSetup-20140330-c314db0-dxva_yuv_range_fix.exe build, and here's some data.

https://docs.google.com/spreadsheet/ccc?...sp=sharing

The numbers listed are the fL measurement from my colorimeter. The images I used are the ones from the colorimeter manufacturer's test DVD. The DVD/MKV obviously refers to the source used, then the renderer used, and then whether or not the limited colour range option was ON or OFF. And then the B60-C40 entries were done with brightness at 60 and contrast at 40.

The results would seem to indicate there are some bugs in there that should be taken care of. There doesn't seem to be any surprises in the first batch with LCR off and brightness/contrast at 50/50, except for the grey-10 results. But I'm wondering if that small error is just due to the encoder possibly making a small change to that image versus the source. I'm thinking it is probably safe to assume that the decoder isn't making any errors. Both the DVD and MKV results for LCR off and 50/50 seem to all be within a margin of error of each other.

But then we start to see problems when we turn LCR on. The DVD software renderer seems to ignore the setting, and the results there are the same as the LCR off results. The rest of the renderers seem to agree with each other, though. And the same goes for the MKV software renderer, and the rest also agree.

And then there are some more disagreements when we move to LCR off B60/C40, and LCR on B60/C40. Again, we see the DVD software renderer showing different numbers than the rest of the renderers. However, we also start to see notable differences between the shaders, DXVA, and DXVA-HD renderers. Brightness/contrast changes do not appear to be getting applied in the same manner for all the renderers.

I haven't had the chance to see what's different with this 2nd build from that day yet. XBMCSetup-20140330-2a56a78-dxva_yuv_range_fix.exe

[jmarshall]

Awesome work!!

I've made a few changes to layout, and have added some comments - it's editable by anyone who wishes to here:

https://docs.google.com/spreadsheet/ccc?...sp=sharing

By the looks there's a couple things worth noting:
1. Software renderer doesn't change at all with limited colour on. i.e. it's not applied at all. (Pink-ish highlight on sheet) This agrees with the code.
2. There's small colour differences based on the source (dvd/mkv) due to colourspace BT601 vs BT709, this is expected.
3. DXVA seems to respond differently to the brightness/contrast stuff (Blue-ish highlight on sheet)
4. There's an odd, unexplained line in yellow. Maybe wrong values here?
5. There should be no difference at all (other than BT601 vs BT709 inaccuracies) in the software renderer between mkv and dvd. There does seem to be large differences - it would pay to check these, as they're the baseline.

The difference with version 2a56a78 will be interesting. Only bother testing DXVAHD here, as that's the only thing that changed. Basically I dropped the DXVAHD brightness/contrast control, and also dropped some limitedcolour range stuff.

Cheers,
Jonathan

[_Shorty]

I don't think there are any data entry errors. All the data entered looks like it matches with what I wrote down on paper while doing the tests. Whenever I got a number that seemed to be out of place I would take more than one reading to make sure it wasn't just a measurement error, but the result would come back the same, perhaps with just a very small margin of error difference. Meaning, usually just the 0.00x would be different. Going to look at the newer build with DXVA-HD right now.

[_Shorty]

ok, seems like that 2a56a78 build is broken for DXVA-HD output, actually. No changes with any run. LCR doesn't make any difference. Brightness/contrast doesn't make any difference. I checked DXVA to be sure, and the changes show up with that renderer just fine. But DXVA-HD is not showing any differences. Added numbers to the bottom of the spreadsheet.

[Piers]

Have the changes from your build been merged into the nightly builds? I think it's important to keep this post updated as much as possible as colour, brightness and contrast accuracy are crucial to any video player.

[Ed76]

hi
im using gotham beta 3 and i am also confused about the "limit color" option: i have a led-tv and activated the limit color option in ati catalyst control center. i test my brightnes with the avshd 709 testvideos. i choose the first one (black clipping) to test my brightness settings and if the "limit color" option works (otherwise the whole screen stays completly black, you cant see the the flashing lines from grey to black).
my question is should i also activate this option in xbmc although i have limit colors activated in ccc? from my test i can say that it seems with activate this option in xbmc the whole screen gets brighter, especially the gui. if i diable limit colors in ccc and activate limit colors in xbmc the screen gets also brighter but in testvideo the screen remains completly black (but a brighter black). so to me it seems this option only change the brightnes or gamma but doesnt limit anything related to colors?

[_Shorty]

I would have to say it is probably best to only use such an option in one place, otherwise changes get additive.

[Piers]

It really depends on:

1) What your TV is capable of (e.g. some Samsung models only allow full RGB ranges on certain modes)

2) What your setting is in CCC.

3) What looks best to your eyes.

An example of this would be using full RGB in CCC and then settings the Limited Colour option in XBMC - this means that Windows will use the full range of colours available (assuming your TV panel supports that) and XBMC will use the --not currently working or sort of working-- colour range that conform to specifications.

Blu-Ray will use, if I'm not mistaken, BT709 and the same for most HD content (including iTunes HD stuff). _Shorty is more knowledgeable in this area but I'm also a major fan of having correct colour, contrast and brightness settings to media is displayed the way it's meant to be.

All said and done it's really what looks best with your eyes - if you notice crushed blacks then have a fiddle with the settings, but don't start changing things too much.

[crawfish]

I switched to Gotham in order to use the 16-235 feature, and it is working fine for me and calibrates consistently with my BD player. I definitely do NOT get the elevated blacks across the board for using 16-235 as I see in the spreadsheet. I have an Nvidia card, Calman Enthusiast, an i1D3, and I can run tests and provide additional data if that would help; just let me know what you need.

I'm currently using Beta 3 with the following settings, and just as a quick test, I switched between all the Render Methods to see how they behave when playing the AVS HD 709 Black Clipping pattern.

Allow hardware acceleration (DXVA2): Off, due to an apparent Nvidia bug which is what led me to switch to the Gotham alphas in order to use the new video levels option as described here:

http://forum.xbmc.org/showthread.php?tid=180884

Decoding Method: Hardware accelerated

Render Method
Auto Detect - fine (this is the setting I use)
Software - crushes blacks (no bars visible in pattern)
Pixel Shaders - fine
DXVA - fine
DXVA-HD - elevated blacks (all bars visible in pattern)

[patseguin]

This seems to have put a bandaid on the problem, good work. Is the dev team incorporating it into future builds? This one says 3/30 so I assume it's beta 3, although it doesn't show the Gotham beta logo on startup.