ATI Radeon HD 5800, 5700 and 5600 Series Thread: Supporting HD Audio Bitstreaming! - Page 176

PDVD8 always looked the same as TMT3 for me, don't know about 9. MPC-HC was always independent, it never gave me trouble. The problems I was having with the 4000 series were inconsistency between PDVD9 and TMT vs MC, and HD vs SD. Different drivers moved things around, but never achieved consistency. I had to do a couple registry fixes for this.

9.12 fixes the problem for HD4000 graphics, and I do get consistent playback in both MPC-HC and powerdvd.

Greetings cyberlolo.

Interesting results with you 4850.

With 4670 and 4550, BTB/WTW is only preserved selecting YCbCr, but not Full RGB.

I guess different card behaves differently....

I have the same experience. With both HD4850 and HD5770, I only get BTB and WTW with 0-255 output when choosing YCrCb 4:4:4

Greetings sonny.

That's why cyberlolo's result is interesting.

Perhaps we'll get a clearer picture after Andy tests his 4xxx cards.

It's not only Cyberlolo, andy also got the same result with his HD5XXX which surprises me a bit.
What display device do you have? I used Panasonic AE-3000 projector.

MPC-HC was never a problem for me, the problem was consistency between MC and PowerDVD/TMT and sometimes SD vs HD. I did get it to work though but that required an MC registry hack.

I was just gonna test the 4550 to see if YCbCr output was actually working, which cyberlolo confirmed, so I don't see a point now. I think the reason you guys are getting that with YCbCr output is that maybe it compresses everything to 16-235 as jong1 proposed a few posts above. So in AVIVO, Full will give you actual 16-235 which then your display has to take care of, and Limited will compress it twice, so after your display expands once, it will look like 16-235.

The 4x cards may behave differently from the 5x cards because of the lack of Dynamic Range option.

I use a Philips 9703 LCD TV calibrated for video level input. It accepts both RGB and YCbCr but there is no user selection.

This is what I get with my HD4850, using RGB Full and madVR configured to full range levels:

I would very much like to know if MadVR can output BTB and WTW w/o gamma correct?

If you set madVR to "TV Levels", it outputs BTB and WTW unmolested, as you can see in my post.

For looking at Desktop levels (non-video) you may find these useful.

 

Black and white level images.zip 120.509765625k . file

Weren't you using "full", or is that what you mean with "TV levels"? Also, isn't madVR unaffected by the Catalyst setting?

Yes, madVR is unaffected by the Catalyst Avivo Video settings, but it's affected by the Catalyst brightness/contrast/gamma normal color settings.

In the other hand, in madVR you have to set "TV Levels" to have Full Range (0-255). If you set it to "PC Levels", then the renderer adjust it's internal brightness/contrast controls (which you can't change by now) so it outputs at 16-235.

The problem is that if you want to have the output at i.e. 16-245 to have wtw, you have to choose TV Levels (Full Range) in madVR, and then apply ffdshow's levels filter to have 0-16 as total black, and 245-255 as total white. This isn't perfect as it introduces banding (ffdshow's levels filter is 8-bit), although I've been told it doesn't matter so much for movies and TV shows.
The perfect solution will be a new version of cr3dlut which will create custom 3dluts for making the trick without introducing banding, but we'll have to wait some time until that new version is released.

Other solutions:
- Using Catalyst brightness/contrast: it decreases PQ.
- Using the TV brightness/contrast controls: but as I've said, I have a KRP-500 (like you), and in that TV, if you use Pure Mode (the best so far) then all the sources share the same settings. So if i lower the brightness to hide 0-16 levels, then that brightness setting is carried over every other source, which I don't want as it makes the picture too dark in the other sources. In the future, when I ISF calibrate the TV, then it will be possible (and the best option too), because with ISF it's possible to have independent settings for each source.

So, for now, the only way to make it is: madVR to TV Levels, and ffdshow's levels filter to calibrate at 16-245. And live with the banding issue until new version of cr3dlut is released...

Andy, I'm not sure we can read anything into your desktop images, since your camera could be exposing things slightly differently between shots. I think we need calibrated images as in my zip above, to be sure what is happening. I think then you may see you get the same video AND desktop levels between RGB limited and RGB full.

This is what I am seeing here:

- RGB Full and Limited seem to output the same desktop levels. The driver (10.2) just seems broken. They both output what should be RGB Limited, i.e. the desktop is compressed so RGB "0" becomes RGB "16".

- YCbCr, as you would expect/hope also compresses desktop levels so RGB "0" becomes luma "16". This works!

- As mentioned before, using limited dynamic range seems to compress video luma 0-255 -> 16-235 (I admit I haven't checked whites, but I'd be surprised if they were not compressed).

- Full dynamic range does nothing to the video levels itself (but see below).

- Using Full DR in combination with RGB (effectively limited regardless of setting) causes black crush. BTB is lost and video black becomes luma "0"

- Using Limited DR with RGB (Limited) outputs consistent levels with desktop. Video black is luma "16" and BTB is restored.

So it seems that when RGB is used video (HD DXVA video?) is always expanded. However, by using the "Limited DR" option this effect can be reversed by first compressing it, before it is subsequently expanded.

I came to this a bit late and have not digested what all of you have been saying. I've only really digested Andy's info! But it seems to me that only YCbCr works as it should (with DR set to full). I can also get consistent levels for desktop and video (but maybe only HD, I have not tested SD) by setting "Limited" RGB and "Limited" DR, but this seems like two errors cancelling each other out to me. It looks like, when RGB is set, video (HD? DXVA?) is always treated as though Full RGB is selected, but RGB in fact is always working in "Limited" mode!!

Desktop aside, Andy, I think all of this is consistent with what you are seeing if your display clips <16, as it seems to.

OK, just to report my test.
I've tried both MadVR and default powerDVD renderer:

YCrCb 4:4:4 and 4:2:0 gives me the same range which is 0-255 if I used power DVD. However as I mentioned the colour is off. When I say the colour is off I don't mean it's unwatchable as Andy shows, it is watchable, but the output luma for red and green are wrong (you need green and red filter in order to see this).

RGB full and limited gives me the same range which is 16-235, and if I set AVIVO to 16-235 or set MadVR to output TV level, I get BTB and WTW, BUT the output range is still 16-235! That is too say, ATI map 0 to 16 and map 255 to 235 (I call it compression).

In all pixel formats, the desktop is compressed to 16-235, and only when you use powerdvd and choose YCrCb you get 0-255.
Jong, our observations are the same. In our case, we still get WTW and BTB, but the range is compressed from 0-255 to 16-235. Anyway, YCrCb seems to be a even worse choice because red and green is not accurate.


If you can see WTW and BTB, there is an easy way to check if you get 16-235 or 0-255: reset brightness and contrast in your display to default and choose to accept 16-235. If you see something more than 16-235, that will tell you that output is 16-235 instead of 0-255.

I do have filters and my colours are identical from YCbCr to RGB Sounds like a wrong colorspace conversion in your display, maybe?

Could be, what display device do you use?
The green and red is a bit off in my case (about 10-15 luma off), I used luma meter in my projector, so maybe it's not noticable if you use colour filter.

Sony Bravia X-series.

If it is so close to identical I cannot spot it with AVS-HD and filters I will live with it (if it exists)! I was using chapters 4&5 of the AVS-HD misc settings. What calibration disc were you using - HD? SD? What were you playing back on? What filter graph? If you were using an SD disc and upscaling in software are you sure you are not suffering from the common problem of converting YUV to RGB using BT.709 (because it is HD when it is converted) even though it should be using BT.601 (because it was SD before upscaling)?

I used AVSHD test patterns, and I used the first video in MISC section. I played it as ISO with PowerDVD, so it shouldn't be SD colour conversion problem.

Another thing I found: only powerDVD output 0-255, I tried MPC-HC with FFDShow, powerDVD9 decoder and almost all renderers, and none combination gives me 0-255. PowerDVD must use some shortcut ATI made for them I think.
With my old 4850, there is no colour issue, and I get identical image if I use Overlay as renderer in MPC-HC.

So just to make things clear. Are you saying that with an HD4850 card, the only way to watch videos with the highest PQ is using PowerDVD and YCbCr 4:4:4? If it's so, I don't understand how with Full RGB (then, if you're right, the video levels are compressed), I'm able to see all the growing scale from 0 (total black) to 255 (total white).


Why? I don't understand this either.


I don't. I get a worse image using Overlay than using evr or madVR. That's for sure.

If you're right and the card outputs 16-235 with Full RGB, how does that affect to PQ? Less dynamic range?

how do you verify this? with what pixel format?
the only consistency I get with the levels.mp4 and avs calibration bluray is when I use YCbCr 4:2:2

Although there can be very slight differences of exposure due to the lens' automatic aperture, the differences are unnoticeable unless you're doing something like stop-motion animation (the reason why for instance, Corpse Bride was done with Nikon manual aperture lenses on a Canon digital SLR). The exposure differences are nowhere near to account for the difference/compensation of 0->16 values. Just eyeballing it, I figure the difference required would be higher than 1/3 stop which is the minimum exposure step in most digital cameras. Even by the lowliest camera standard, it would be useless if it couldn't resolve such a difference with its aperture timing.

Besides that, all of the pics that are expected to look the same given the settings, look the same (like 1.1.3 and 2.2.3). The ones that are expected to be compressed (1.2.3) or clipped (2.1.3) also show it. The likelihood of the desktop levels really being off but by random exposure differences somehow compensating to show as expected, in all of those individual cases, is very slim to say the least.

Exposure differences, more importantly, cannot compensate for compression/expansion, only for overall brightness of the whole range.

I just used regular pics to be compared to each other, cause there are only 3 scenarios possible, correct, clipped, or compressed. I can see all three in some examples above just as expected, so I just used pics with detailed dark areas. They are not meant to be compared to the video pictures, I just wanted to know that the desktop was being changed at all.


What type of input are you setting your display to? If you set the input of your display to 0-255 input, do both pixel format limited and full in CCC look the same?


I think this could be due to your display not accepting full 0-255 signals, but that wouldn't be a problem with the card.


It only clips when input is set to 16-235 or YUV though. With 0-255 input (section 1 in my post) it's not clipping anything. It seems to me that your and Somy's observations are consistent with your displays not taking in 0-255 properly?

I confirm the same result with no EDID override in my setup (ATI HD5850, Denon AVR-590, Dell U2410), just catalyst 10.2 (ATI HDMI driver). I tested playback on PDVD9 (2528) and TMT3 (170). Well, we really have something new in the cat 10.2.

OK, I am confused.

If I want the best possible image sent to my TV, what output should I use?

This is assuming my AVR and TV can accept all four types (full RGB, limited RGB, 4:2:2, and 2:2:2).

Also assume everything works as it should - I understand the current debate is over things working properly or not.

Hmmm....I CANNOT confirm with my Denon AVP A1...no True HD unless I have the EDID override. Brand new Windows 7 64 bit, Catalyst 10.2, ATI drivers, (no Realtek), TMT3 .170 beta....

Does PDVD install "something" that makes this work? I actually own PDVD but do not have it installed on this latest OS build, in an effort to keep things as minimal as possible.

TrueHD too? And do you really get the third option on PDVD9? Could you double check? Like thrang, these fail for me without the override with my Denon 988. Thx.

Actually PDDVD9 is the hardest, for you either can bitstream both DTS HD & TrueHD or you bitstream nothing. With other players you get at least DTS HD if you need the override but haven't installed one.