PDA

View Full Version : Any success with Rec. 2020 10-bit?


benwaggoner
4th June 2014, 18:58
So, because of my unslakable curiousity, I'm trying to figure out a way to see Rec. 2020 10-bit. I'm encoding from a 1080p59.94 10-bit DNxHD source to two Main 10 files identical but for

2020: --colorprim bt2020 --transfer bt2020-10 --colormatrix bt2020nc

709: --colorprim bt709 --transfer bt709 --colormatrix bt709
I then mux them with a daily build of mp4box.

I've got a wide gamut Dell monitor with a 10-bit playback path from a K4000 GPU, and a 2020 .icm file installed in Windows. I've got a 2014 Samsung UHD TV which can decode HEVC 10-bit and supposedly can do Rec. 2020 10-bit.

VLC plays back the frames corrupted (10-bit issue?). MPC-HC and WMP play back perfect. After Effects won't even import them.

With everything that will play back, the two files look identical. Is this because everything ignores the Rec. 2020 flag, or because it isn't getting muxed in correctly? In MediaInfo, I don't see any color space data at all:

Video
ID : 1
Format : HEVC
Format/Info : High Efficiency Video Coding
Format profile : Main 10@L4.1
Codec ID : hvc1
Codec ID/Info : High Efficiency Video Coding
Duration : 2mn 0s
Bit rate : 10.1 Mbps
Maximum bit rate : 14.1 Mbps
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate mode : Constant
Frame rate : 59.940 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 10 bits
Bits/(Pixel*Frame) : 0.081
Stream size : 144 MiB (100%)
Title : 265@GPAC0.5.1-DEV-rev5261
Encoded date : UTC 2014-06-04 17:35:32
Tagged date : UTC 2014-06-04 17:35:33

Any insights/suggestions?

vivan
4th June 2014, 20:02
MPC-HC and WMP play back perfect. After Effects won't even import them.

With everything that will play back, the two files look identical. Is this because everything ignores the Rec. 2020 flag, or because it isn't getting muxed in correctly?Are you using madVR or EVR/something else?
Because only madVR respect those flags, other renderers ignore them.

And if you're using madVR - check what OSD (Ctrl+J) says.

huhn
5th June 2014, 02:40
you have to say MadVR first that your display is bt 2020 but if you do that it should work but the result is "only" dither 8 bit rgb

kolak
5th June 2014, 12:18
Has your source been graded to Rec.709?
If yes I assume you would see 0 difference regardless of the setting.
You would need source from some high-end, wide gamut camera (like F65) and than convert it straight from 16bit EXR/TIFF etc.

Samsung TV does support Rec.2020 as long as you turn it on in the settings and when signal comes from HDMI.
It's not clear if it will use it when playing x265 file from USB.

Both displays can do more than Rec.709, but they are no near of covering whole 2020 gamut. Proving anything can be very difficult :)

I would use 16bit source file, than put it through fmtconv (easier with vapoursynth than avisynth), which does support Rec.2020 and than encode it as 10bit x265.

benwaggoner
5th June 2014, 18:59
Has your source been graded to Rec.709?
If yes I assume you would see 0 difference regardless of the setting.
You would need source from some high-end, wide gamut camera (like F65) and than convert it straight from 16bit EXR/TIFF etc.
I'm using R3D sources, applying a Rec. 2020 working space in After Effects, and yanking up the saturation to something below Rec. 2020 clipping, but out of gamut for Rec. 709. Then rendering out to a v210 AVI and encoding from there. Since x264 and x265 ignore the source profile and just take in the values, flagging as 709 and 2020 from that source should yield very different playback results, with the 709 a lot less saturated (but still pretty saturated)>

Samsung TV does support Rec.2020 as long as you turn it on in the settings and when signal comes from HDMI.
It's not clear if it will use it when playing x265 file from USB.
Ah, that might be it. Sounds like it's MadVR + laptop time.

Both displays can do more than Rec.709, but they are no near of covering whole 2020 gamut. Proving anything can be very difficult :)
Well, I'd like to see more saturation, particularly in the high luma, in the 2020 than the 709 clip.

kolak
5th June 2014, 21:12
Please post results for your experiment.
Shame madvr can't do 10bit.

You can try Tweak RV player if you have quadro card. It should be able to send 10bit over display/hdmi port as it can be set to work with 10bit. I don't think it will play h265 thought (it does h264 through ffmpeg engine).

benwaggoner
5th June 2014, 21:29
Please post results for your experiment.
Shame madvr can't do 10bit.

You can try Tweak RV player if you have quadro card. It should be able to send 10bit over display/hdmi port as it can be set to work with 10bit. I don't think it will play h265 thought (it does h264 through ffmpeg engine).
I do have a Quadro K4000 card. I could play out a 2020 H.264 file. I'll give that a shot.

huhn
6th June 2014, 00:07
the problem is you need a surface that is 10 bit. every normal gpu can output 10 or more bit over hdmi/dp for years on a dx 11 surface. there is simply no software that use that. i only know about opencl programs like photoshop where a card you use is need for that that really uses 10 bit.

benwaggoner
6th June 2014, 20:12
the problem is you need a surface that is 10 bit. every normal gpu can output 10 or more bit over hdmi/dp for years on a dx 11 surface. there is simply no software that use that. i only know about opencl programs like photoshop where a card you use is need for that that really uses 10 bit.
Premiere Pro CC will also output OpenGL 10-bit, over DisplayPort at least. I could use a DP to HDMI converter perhaps. Convincing Windows/Premiere to output using Rec. 2020 will be another challenge.

huhn
6th June 2014, 20:55
it is possible to force EVR to output 10 bit rgb and a option to force 10 bit rgb input. but in the end windoes will turn it in 8 bit again.

i also saw in the geforce rgb about 10 bit that this disables aero so it looks like aero can't do 10 bit at least in windows 7 hope this info doesn't matter but how knows.

and the next thing is, is 10 bit rounding better than very high quality 8 bit dithering? it has to be transformed to rgb so this matters.

maybe using MadVR with "this device is already calibrated" with bt. 2020 is the easiest way of cause with a bt 2020 icm or a bt 2020 3d lut. MadVR can deviantly deal with bt. 2020 and can transfrom it for normal displays to bt. 709 only 10 bit is missing. and if i'm not totally mistaken it can use file names to change the colormatrix so if the stream doesn't have these information you can add them this way.

xooyoozoo
8th September 2014, 07:48
4K Blu-ray is coming soon with 10bit BT.2020 HEVC. (http://www.cnet.com/news/4k-blu-ray-discs-arriving-in-2015-to-fight-streaming-media/)

It's certainly nice to have mass-market 10bit content, and maybe the scramble to make cheap 10bit HEVC decoders will spread those designs into other markets too.

(Too bad they didn't decide to also bump up chroma subsampling while upgrading everything else)

Sparktank
8th September 2014, 08:54
I was going to create a separate thread for that news in the general section.

I'm curious about the new discs coming out.
With support up to 60fps for content, I wonder how that will work for movies filmed in 48fps, like the Hobbit trilogy.
I'm not interested in anyone's opinion on the soap opera effect, I've heard plenty on the internet by now.

I hope HEVC will be maturely developed by then. I imagined it would be quite efficient by the time they stat marketing the 4KBD's.

Another set of remasters is definitely going to worth looking at.
Maybe, we'll get a real Silent Hill remaster without importing.

STaRGaZeR
8th September 2014, 17:55
Still 4:2:0, goddammit! Still, nice improvements with BT.2020 and 10-bit.

SeeMoreDigital
8th September 2014, 18:04
4K Blu-ray is coming soon with 10bit BT.2020 HEVC. (http://www.cnet.com/news/4k-blu-ray-discs-arriving-in-2015-to-fight-streaming-media/)

Indeed it is coming... This was confirmed to me on Saturday at IFA by Ron Martin of Panasonic Hollywood Laboratory (PHL). ;)

sneaker_ger
8th September 2014, 18:20
They also announced the "Ultra HD" specification and logo at the IFA. Unfortunately the lowest denominator is BT.709 (with 10 bit, though), reserving things like BT.2020 and 12 bit for a future second generation UHDTV. They also said "HDR" for the second generation but I'm not sure if that's something different than BT.2020.

xooyoozoo
8th September 2014, 23:06
Unfortunately the lowest denominator is BT.709 (with 10 bit, though), reserving things like BT.2020 and 12 bit for a future second generation UHDTV.

For the immediate future, I don't see 709 as a bad thing. Because so many displays can barely even do 709, a 10bit 2020-remastered release would waste a chunk of the available bit depth on essentially unavailable spectrum. (Edit: "wasted" because while 709 is happy with 10bit, 2020 needs more to prevent banding)

They also said "HDR" for the second generation but I'm not sure if that's something different than BT.2020.

There's been some work done on perceptually-aware, float-esque mappings between integer pixel values and optical values (here's Dolby's work (https://www.smpte.org/sites/default/files/2014-05-06-EOTF-Miller-1-2-handout.pdf)). By definition, the 'HDR' feature wouldn't be a part of Rec. 2020, but it could use the same chromaticities and 12bit colordepth, and still scale scale to 10k nits without banding.

benwaggoner
9th September 2014, 00:47
Still 4:2:0, goddammit! Still, nice improvements with BT.2020 and 10-bit.
Chroma subsampling is going to matter even less at UHD. It's the same chroma as 1930x1080 4:4:4, and I've never even seen a synthetic test pattern that showed visible chroma artifacts with that.

STaRGaZeR
9th September 2014, 15:26
I did, as have a lot of people, but this is already discussed to death. There will always be artifacts no matter the resolution. This is just hardware vendors pushing for cheaper solutions. I can only imagine the discussions about the absolute need for 10-bit in their HEVC decoders and how much money and die area they would have to spend on it.

Only increases in pixel density can somewhat negate the effects of chroma subsampling.

foxyshadis
10th September 2014, 00:41
I already don't spot 2x2 chroma sampling at 720p and up, except for artificial tests; for 4k it could be 4x4 subsampled and I doubt I'd notice or care. (Instead, we can just more aggressively quantize it with HEVC's huge blocks.) Same with depth, as density increases dithering becomes more indistinguishable from a real depth increase, so using it for wider gamut makes more sense.

As long as the new standards allow 10-bit 709 in SD and HD, 10-bit 2020 UHD isn't a bad trade-off (equal to a bit more than 9-bit 709). 12-bit is still exotic, 10-bit is at least proven and widely implemented now.

STaRGaZeR
10th September 2014, 11:06
That's the thing: quantize it instead of subsampling it. Well, maybe for the 8K iteration!

zerowalker
11th September 2014, 00:39
Will this mean that things may be recorded in 10bit in the future?
And possibly remastered in 10bit (non-digital content), as i guess that should be possible as Analogue doesn't have a definition of bit depth.

benwaggoner
24th September 2014, 19:14
Will this mean that things may be recorded in 10bit in the future?
And possibly remastered in 10bit (non-digital content), as i guess that should be possible as Analogue doesn't have a definition of bit depth.
Stuff has been produced and recorded in 10+ plus for YEARS now. Going back to D5 and HDCAM-SR tape.