AVS Forum banner
Status
Not open for further replies.

16-235 0-255 my turn!

8K views 15 replies 7 participants last post by  Li On 
#1 ·
Hi,


Sorry for the repeat post. All other similar topic gives me headache! :D


IMO, there is INPUT level (DVD, software MPEG2 decoders) and OUTPUT level (Overlay. VMR, display card output).


For normal PC operation (Word, PhotoShop, Web) I guess both INPUT and OUPUT are at 0-255 level.


DVD is suppose encoded in video level which is in the range of 16-235. PC desktop has all the value in 0-255.


First question, how's software MPEG2 decoders (WinDVD, PowerDVD, Cine, NVDVD) treats DVD content? As 16-235 or 0-255 level?


If the MPEG2 decoder process DVD as 16-235, then where is 235 data? Throw away?


Does video renderer (Overlay, VMR) remap/expand 16-235 to 0-255? I guess not because not all video content are DVD MPEG2. If video renderer does level conversion then other non DVD content will have problem.


Then how's the 16-235 fit to the display card OUTPUT level? All PC display card output in 0-255 level, right? I don't see there is a option for 16-235 level output in display property. And if so then normal desktop will be affected.


So how all these add up? Personal I think in HTPC world everything is treated as 0-255. I have no problem with these as we get everything from DVD.


I always get the 2 moving black bars and 2 moving white bars in AVIA.


In the 2 moving black bars pattern, I guess the brighter (right) bars is at 7.5 IRE (higher than 16), the darker (left) bar is at between 0-16 and the background is at IRE 0 (for 16-235 player, it means 16, for 0-255 player, it means 0, I guess). If the player (software MPEG2 decoder) works in 16-235, then the darker bar will be invisible.


All in all, it's not a issue for me. And I don't want to think it anymore as it still cause me headache!


regards,


Li On
 
See less See more
#2 ·
Quote:
Originally posted by Li On


I always get the 2 moving black bars and 2 moving white bars in AVIA.


In the 2 moving black bars pattern, I guess the brighter (right) bars is at 7.5 IRE (higher than 16), the darker (left) bar is at between 0-16 and the background is at IRE 0 (for 16-235 player, it means 16, for 0-255 player, it means 0, I guess). If the player (software MPEG2 decoder) works in 16-235, then the darker bar will be invisible.

In Avia both black bars are above black (16) and the white bars are below peak white (235)
 
#3 ·
Hi Jeffy,


Thanks! How about list the exact value and IRE level of the moving black and white bars?


moving black bars pattern

- background (value, IRE)

- left (dark) black bar (value, IRE)

- right (bright) black bar (value, IRE)


moving white bars pattern

- white background (value, IRE)

- left (dark) white bar (value, IRE)

- right (bright) white bar (value, IRE)


And the FULL black pattern (value, IRE (0 I guess)). The FULL 100 IRE pattern (value 235?).


And you're saying there is no 235 data in the AVIA?


Thanks in advance.


regards,


Li On
 
#4 ·
I posted this in "the other thread" and thought it might help here, too. Throw some numbers in and see what happens.


If DVD's are mastered into YCbCr from studio RGB space (as I suspect they are), then some clipping may occur upon playback in the PC RGB space. Also note that it is possible to plug in YCbCr values that result in clipping in the studio RGB space, but if the source was originally studio RGB, then those values wouldn't be present in YCbCr unless additional mastering was performed while in YCbCr space.

 

component to rgb conversion.zip 2.0498046875k . file
 

Attachments

#5 ·
If I wasn't about to walk out the door, I'd throw together another sheet to convert all studio RGB (including BTB and WTW) into YCbCr, and then convert those YCbCr values to PC RGB to see which ones (if any) are clipped. Then I'd have a better idea of the magnitude of the problem.
 
#6 ·
I was worried about this stuff for a while, but not any more. I did try all possible input-output clipping combinations, and none of them look better than not messing with the levels at all (Nvidia drivers with Nvidia Forceware Multimedia 3.0 Beta). Whatever happens elsewhere in the path (VMR9, video card drivers, decoder) is out of my control, so why worry. It sure looks great as is. I did have some issues with ATI at one point where something was wrong somewhere, but at one point driver upgrade did the trick.


That being said, color space conversions for DVD playback are multiple. This is an example of DVD with FFDShow playback in my setup with color spaces and bit depths. Again, not sure what happens with levels (i.e. video range).


1. DVD native into decoder: YCbCr (8bit)

2. Decoded Stream into FFDShow: YV12 (12 bit, but due to its different characteristics and with unpacked chroma, is equivalent of color depth of RGB24, YUV2 for other players)

3. FFDShow processing and output to video card: YV12 (can force other)

4. Video card conversion to desktop: RGB32 (8-bit) (this is important step because this step upconverts 4:2:0 to 4:2:2 format and if done in hardware, is chroma-bug-free)

5. Video card internal processing: expanded colorspace (128-bit floating point precission for) VMR7/9, RGB32 (8-bit) for overlay

6. Video card conversion to output (for DVI only): RGB24 (8 bit, just discard opacity, should not really be a full fledged colorspace conversion)


Here is a good link to MS library that discusses some of these issues:

http://msdn.microsoft.com/library/de...YUVFormats.asp
 
#7 ·
Hi,


I think once the picture left the MPEG2 deocder, the path to display card output should be all in 0-255 PC level. Unless you change the level in ffdshow.


The question remain the process of MPEG2 decoder (in software decoder mode I mean). Whether the MPEG2 decoder:


a: decode 0-255 and output 0-255

b: decode 0-255 and output 16-235 (compress) as video level

c: decode 16-235 and expand output at 0-255

d: decode 16-235 and output 0-255 with 16-235 data

e: decode 16-235 and output 16-235 as video level


I guess b and e is not possible because I don't think the processing after MPEG2 decoder could understand video level.


I always thought the AVIA black and white moving bars pattern has 235 level. As I can always get the 2 pattern working at the same time, I think the MPEG2 decoder works in "a" mode, decodes 0-255 and output 0-255.


But Jeffy said the 2 AVIA patterns are still within 16-235. So I need other test pattern to check. Any suggestion? Thanks in advance.


regards,


Li On
 
#8 ·
As far as I know DVE's sub-black is in 0-16 range and sub-white is in 235-255 range. You can get all of the bars on grayscale adjustment screen in 5% increasements.


There should also be a section (under special features I believe) in all THX discs with THX patterns. I did not use that in a while, but they should also have sub-levels.


Overall, I'd say if you cut input to 16-235 and map output to 0-255, if you don't see a bar, it should be in sub-ranges. If you still see it, it is probably set at 16 or 235, so try cutting 17-234 to confirm.
 
#9 ·
Hi,


Just tested the DVE Title 12, Chapter 14 "Reverse Gray Ramps" pattern.


My system: original Radeon 8500 VGA out, ZoomPlayer Pro 4 Beta 1, NVDVD 3.0 (leak beta) with VPP, ffdshow (input YV12 from Info, output selected only YV12), Shapen and Picture propers (no resize), Overlay.


At default color control, I get all the levels and and between the "DOTS" range. But NOT the WTW and BTB! Play with the ffdshow Picture properties Gamma setting, I can get ALL the WTW and BTB, but not both end at the same time! Switch to YUY2 output only with same result.


That prove the NVDVD MPEG2 decoder decodes and outputs ALL 0-255 to ffdshow. But all the levels can be output at the same time via Overlay.


I can't do VMR9 as it has heavy tearing. So I tried VMR7 (with a Dscaler running in the background to get the Overlay first!) which still has a tiny bit of tearing but still watchable.


Well, well, well, with VMR7 output (still in YV12), at default color control, ALL the BTB and WTW are visible at the same time. The whole ramp bar is smooth end to end.


So I tune (heavy push of Sharpen etc) the overall VMR7 picture to match my usual Overlay looking. So far so good! I checked a few clips and "BLACK" detail is indeed better. Also there seems a bit more "WHITE" detail! Some dark banding problem is indeed much better now!


I need to watch more...


regards,


Li On
 
#11 ·
These issue are important when viewing DSS on a projector calibrated on a PC.


It is best in this case to adjust both the white and black levels (for the 7.5 IRE+scale change) to compensate. Otherwise the picture will look "washed out". The blacks not black enough and the whites not a full white.





Joel
 
Status
Not open for further replies.
You have insufficient privileges to reply here.
Top