Setting the colorspace to YCbCr with a range of 16-235 in Nvidia Control panel seems to have done the trick, as both VMC and Arcsoft are now passing BTB and WTW info.  I don't have the DVE disk yet, but I could immediately tell that there was better detail in the black level of Arcsoft.

 

---

XPS420 w/ 2 ATI OCURs
HDHomerun
DMA2100 Extender x 2

Montyward:

Setting the colorspace to YCbCr with a range of 16-235 in Nvidia Control panel seems to have done the trick, as both VMC and Arcsoft are now passing BTB and WTW info.  I don't have the DVE disk yet, but I could immediately tell that there was better detail in the black level of Arcsoft.

Are you sure it is actually passing BTB and WTW info through to the display?

With my ATI experience, BTB and WTW are not passed through to the display.  In my case I calibrate my system so the source range of 16-235 is visible (and not oversaturated) on my HDTV (16 being black, 235 white).  If I try to calibrate so I can see anything (from the source) below 16 or above 235 on my HDTV, I find it is simply not available after video processing (unless I use hacks to both turn off 16-235 -> 0-255 expansion at video processing and turn off 0-255 -> 16-235 compression at output).

Personally I don't have an issue with loosing BTB and WTW info, but I'm sure video perfectionists would.

BTW, if you are looking for Bluray and/or HD-DVD calibration sources, the ones available on this page work well:  http://www.avsforum.com/avs-vb/showthread.php?t=948496

---

STB w/R5000HD USB I/O, Gigabyte GA-P35-DS4, Quad Q6600, 4.0 GB RAM, ATI HD 3870 512MB, Ultra XVS 600W PSU, 3x SATA 500GB, 2x SATA 300GB, LG GGC-H20L, PVR-250, Toshiba 51H83 (51" HDTV), Yamaha RX-V2400 Amp, 5x Energy Speakers, SVS Sub, Harmony 880 Remote

iank:

FWIW the blu ray HQV test disc runs in a third party app hosted in MCE independent of the MCE pipeline. If it failed so many of the HQV tests then the 3rd party decoder they were using has some issues that need to be sorted out. The jaggies tests are on 1080i content, something you will rarely encounter on BRD. For 1080i broadcast content you will use the MSFT decoder and in this case we will get inverse telecine and decent adaptive deinterlace if the GPU supports it and it is exposed in the proper manner via DXVA. MCE is bound by the performance of the GPU and its video process core (both AMD and NVDA have dedicated video processing silicon on their GPUs).

But you are absolutely right, there is a lot more that can be done to improve video quality in MC in general by all parties involved.

I transcoded/remuxed a lot of the HQV test content to run natively in the MCE pipeline. I found that, at least with NVIDIA, inverse telecine of 1080i content worked fairly well although recovery from lost cadence was sometimes slower than one would expect.

With deinterlacing of 1080i video I find that PDVD 7.3 and a recent release of Arcsoft TMT (v125 beta) pass the HD HQV benchmark jaggies test when played from within the Blu-ray container. If however I attempt to play the same jaggies test with PDVD and TMT directly from the .m2ts file, both players fail. In both cases there is a difference in CPU load between GPU HA enabled and GPU HA disabled which suggests to me that HA and choice of deinterlacing can operate independent of one another.

I can replicate the above results with recorded 1080i TV broadcasts. By remuxing the recorded MPEG-2 video into a Blu-ray container I get better deinterlacing in PDVD and TMT than I do with direct playback of the .ts recording.

I use VMC for live TV broadcasts but in the absence of being able to achieve the highest quality of deinterlacing outside the Blu-ray format leaves me with deinterlacing that is inferior to what my hardware can offer.

I'm interested in how I can achieve the HQV standard of deinterlacing outside the Blu-ray format and more specifically if it can be achieved in VMC where the PC hardware supports it.

GPU: Nvidia 8600GTS. OS: Vista Ultimate.

 

 

 

 

 

HT Slider:

Are you sure it is actually passing BTB and WTW info through to the display?

With my ATI experience, BTB and WTW are not passed through to the display.  In my case I calibrate my system so the source range of 16-235 is visible (and not oversaturated) on my HDTV (16 being black, 235 white).  If I try to calibrate so I can see anything (from the source) below 16 or above 235 on my HDTV, I find it is simply not available after video processing (unless I use hacks to both turn off 16-235 -> 0-255 expansion at video processing and turn off 0-255 -> 16-235 compression at output).

Yes, I'm sure since I'm seeing three black bars on a black background on either side of the white pattern.  I can't remember the pattern name.  I think because my TV can be set for PC or TV that perhaps that is why I am getting it?  I really have no idea at this point.

Again though, playing a DVD through Arcsoft resulted in a loss of detail in black areas, so perhaps its back to the drawing board.  Anything that was dark, like dark hair or a dark shirt appeared black and lacked any detail (where I'm sure detail was meant to be seen).

---

XPS420 w/ 2 ATI OCURs
HDHomerun
DMA2100 Extender x 2

iank:

<snip>

But you are absolutely right, there is a lot more that can be done to improve video quality in MC in general by all parties involved.

I transcoded/remuxed a lot of the HQV test content to run natively in the MCE pipeline. I found that, at least with NVIDIA, inverse telecine of 1080i content worked fairly well although recovery from lost cadence was sometimes slower than one would expect.

In your 1080i test your STB has the advantage of outputting at 1080i a 1080i signal. MCE will process to 1080p first, then the GPU re-interlaces to 1080i. So the image will get softer. There is an architectural disconnect on the platform here that currently prevents us from syncing/gen-locking to your display for doing "native" res output.

As far as difference in color, I can tell you that via DVI the color is correct. We process 709 and 601 color space accurately. The color is to spec in the frame buffer. If your connection is YPbPr I can tell you that all GPU vendors have issues with taking the RGB frame buffer back into the component space accurately.

Finally, yes, GPU based video processing is not nearly as good as dedicated silicon. And that is where the real issue lies IMHO. I would love to see anyone implement the Silicon Optix or Algolith technology on a modern GPU.

Ian,

If your still reading this post, I'm wondering at this time has there been any advancement with hardware scheduled for Windows 7 to help in the general picture quality over all, with better hardware that will output a true 1080i image from a 1080i source.

As I have seen here, this needs to be done in hardware and my question is, is Microsoft working with any vedors to build a video card that is good for just HTPC useage ?

---

-Dave

MCP, MCSA, MCSE 2003
Windows Vista Connected Exp:Home Theater for Technologists
Windows Vista Connected Exp:Home Theater for Sales professionals

Home theater specialist (10+ years)

"Have you or someone else at Microsoft brought these issues to the attention of the video card manufacturers?  Do you have a process in place where Microsoft can ensure ATI and Nvidia fix fundamental issues like these that severely take away from Media Center's ability to perform?"

Yes, everyone involved is aware of the issues here.

With respect to things like grayscale and color accuracy: we test for that and file bugs as we find them. It is easy to test with 601 and 709 colorbars as well as a nice ramp pattern. The level expansion issue you saw was a driver "feature" where they try to be smart, or optimize for the PC display rather than a consumer display. MC defualts to 16-235, reference video levels. Great for your big TV, not so hot for a PC desktop monitor calibrated to 0-255. And things get even stranger on HDMI where spec compliance for the HDMI license forces some compromises on IHVs that result in less than optimal results.

Broader work with video processing improvment requries work by all parties involved. We are cought between the PC and the consumer display and getting it to work right in all cases wihtout resorting to some sort of video nerd menu is defficult. Although I'd be OK with a video nerd menu, personally.

---

Ian Kennedy - eHome TV Playback Test Lead
This posting is provided "AS IS" with no warranties, and confers no rights.
Box:P5B-E,8600GT,E6600,2GB,Seasonic PSU.Zalman HD160. 1x WD SE16 1TB, 1x OCZ 32GB SSD. 2x OCUR /w CableCard.

DavidinCT:

Ian,

If your still reading this post, I'm wondering at this time has there been any advancement with hardware scheduled for Windows 7 to help in the general picture quality over all, with better hardware that will output a true 1080i image from a 1080i source.

As I have seen here, this needs to be done in hardware and my question is, is Microsoft working with any vedors to build a video card that is good for just HTPC useage ?

I really doubt this will happen. The odds are better for getting improved video processing into the pipeline.

At this time the only reason for supporting native video output is for use with an outboard video processor and I don't think the IHVs would be interested in that.

---

Ian Kennedy - eHome TV Playback Test Lead
This posting is provided "AS IS" with no warranties, and confers no rights.
Box:P5B-E,8600GT,E6600,2GB,Seasonic PSU.Zalman HD160. 1x WD SE16 1TB, 1x OCZ 32GB SSD. 2x OCUR /w CableCard.

I see some folks are having fun with the NVIDA HDMI output. Ahem. Here's some info you may find interesting:

1- The HDMI spec and therefore the license requires all video be constrained to the range of 16-235
2- To ensure this and be compliant the driver at the "scan-out" stage when it actually spews the video to the connector scales the framebuffer from 0-255 to 16-235.
3- This sounds OK right?
4- Not really, our video is ALREADY 16-235 constrained so this process hoses it. This is despite our setting 16-235 via DXVA.
5- The work around is to use the NV control panel and force 0-255, which "expands" video to match the desktop/PC range and things then match when "the squeeze" happens. But you lose BTB and WTW in this case. Note that at this time the NV control panel setting for dynamic range is only for video, and that is why this works.

You can preserve BTB and WTW by leaving things alone, or forcing 16-235 in the CPL, but with only 8bits per color channel you're gonna get banding no matter what and the UI and photos will look odd once you calibrate to get black levels right.

This is why I use a DVI->HDMI cable (bluejeanscable.com FTW). The machine thninks its on DVI and all is well. No audio support, but that's OK for now.

And yes, the right people know about this.

-Ian

P.S. Little known fact: if you ever meet the program manager for Media Center playback (his name is Wes) ask him "hey Wes, what's up with that nominal range issue in Media Center" you'll actually get to see someones head explode like in the movie "Scanners." It's awesome.

---

Ian Kennedy - eHome TV Playback Test Lead
This posting is provided "AS IS" with no warranties, and confers no rights.
Box:P5B-E,8600GT,E6600,2GB,Seasonic PSU.Zalman HD160. 1x WD SE16 1TB, 1x OCZ 32GB SSD. 2x OCUR /w CableCard.

iank:

DavidinCT:

Ian,

If your still reading this post, I'm wondering at this time has there been any advancement with hardware scheduled for Windows 7 to help in the general picture quality over all, with better hardware that will output a true 1080i image from a 1080i source.

As I have seen here, this needs to be done in hardware and my question is, is Microsoft working with any vedors to build a video card that is good for just HTPC useage ?

I really doubt this will happen. The odds are better for getting improved video processing into the pipeline.

At this time the only reason for supporting native video output is for use with an outboard video processor and I don't think the IHVs would be interested in that.

Riddle me this, Do you know of any video card out there, that Media Center will output a true 1080i image from a 1080i source with component video output ? This card will need to play back 1080i also HD-DVD/Blu-ray (downcoverted to 1080i)

---

-Dave

MCP, MCSA, MCSE 2003
Windows Vista Connected Exp:Home Theater for Technologists
Windows Vista Connected Exp:Home Theater for Sales professionals

Home theater specialist (10+ years)

iank:

Broader work with video processing improvment requries work by all parties involved. We are cought between the PC and the consumer display and getting it to work right in all cases wihtout resorting to some sort of video nerd menu is defficult. Although I'd be OK with a video nerd menu, personally.

How about documented reg settings.  You guys don't have to worry about usability, but the flexibility is still there for the nerds.

---

DVRMSToolbox, Recording Broker, LcdWriter, and more software
babgvant.com

iank:

I see some folks are having fun with the NVIDA HDMI output. Ahem. Here's some info you may find interesting:

1- The HDMI spec and therefore the license requires all video be constrained to the range of 16-235
2- To ensure this and be compliant the driver at the "scan-out" stage when it actually spews the video to the connector scales the framebuffer from 0-255 to 16-235.
3- This sounds OK right?
4- Not really, our video is ALREADY 16-235 constrained so this process hoses it. This is despite our setting 16-235 via DXVA.
5- The work around is to use the NV control panel and force 0-255, which "expands" video to match the desktop/PC range and things then match when "the squeeze" happens. But you lose BTB and WTW in this case. Note that at this time the NV control panel setting for dynamic range is only for video, and that is why this works.

You can preserve BTB and WTW by leaving things alone, or forcing 16-235 in the CPL, but with only 8bits per color channel you're gonna get banding no matter what and the UI and photos will look odd once you calibrate to get black levels right.

This is why I use a DVI->HDMI cable (bluejeanscable.com FTW). The machine thninks its on DVI and all is well. No audio support, but that's OK for now.

And yes, the right people know about this.

-Ian

P.S. Little known fact: if you ever meet the program manager for Media Center playback (his name is Wes) ask him "hey Wes, what's up with that nominal range issue in Media Center" you'll actually get to see someones head explode like in the movie "Scanners." It's awesome.

Any trick you can recommend for those of us with monitors that only have HDMI input?  Using a DVI->HDMI cable doesn't work, as somehow the card still thinks it's using HDMI.  I'm using an Nvidia 8800GT and LG 246W monitor.

Thanks,

Bryan

iank:

P.S. Little known fact: if you ever meet the program manager for Media Center playback (his name is Wes) ask him "hey Wes, what's up with that nominal range issue in Media Center" you'll actually get to see someones head explode like in the movie "Scanners." It's awesome.

LOL

---

-Dave

MCP, MCSA, MCSE 2003
Windows Vista Connected Exp:Home Theater for Technologists
Windows Vista Connected Exp:Home Theater for Sales professionals

Home theater specialist (10+ years)

Hmmm...  I've built two HTPCs recently, one AMD HD 3200 (now upgraded to HD 4550 for HDMI multichannel audio and better de-interlacing performance) and one nVidia 9400 based.

I come from a broadcast video background - and was amazed at how messed up the PC-world has got when it comes to levels and colour space.

The ATI BT601CSC hack at least keeps black levels and white levels consistent between SD and HD content (I don't massively care about BTB and WTW as an end-user - I would if I was engineering a broadcast chain).  Having BTB and WTW can be useful in calibrating a display though.  I've yet to run a proper DVE calibration on SD and HD on my ATI set-up - but visually post-BT601CSC registry change I at least have consistency between SD and HD sources.  (HOW can ATI/Microsoft ship something so flawed by default?  If Sony shipped a DVD player or a Blu-ray player so massively wrong they'd be a laughing stock...)

I'm still struggling with nVidia (The registry keys posted by iank should be useful I hope)

One other thing that REALLY confuses the heck out of me.  Why any discussion at all about 7.5IRE here (a North American Composite NTSC black levels set-up - not used in Japan) when discussing digital video.  7.5 IRE should have no place in any discussions of non-composite stuff - and is certainly nothing to do with black level being at digital 16 in Rec 601 or 709 stuff.  (The black at 16 / white at 235 rather than 0 and 255 decision was taken to ensure undershoots and overshoots present on analogue sources were preserved through digital paths - clipping them would introduce ringing - you need some headroom above and below to ensure this.)  As a European based consumer - I've never needed to worry about 7.5IRE (my DVD players replay 480i R0/R1 DVDs without 7.5IRE set-up over HDMI or RGB SCART - no black level variation between 480i and 576i titles) 

I have heard that some US HDTVs incorrectly call their "Video" 16-235 level mode "7.5IRE" for some reason... Not sure why this is...

Surely there are two main issues here :

1. Output standard - 16-235 vs 0-255, YCrCb vs RGB (and 4:4:4 vs 4:2:2 for YCrCb)  This should be set by default to 16-235 (and probably 4:4:4 YCrCb if supported, if not 4:2:2 YCrCb?) when an HDMI connection is detected, or 0-255 4:4:4 RGB if a DVI connection is detected. BUT the user should be able to over-ride this - and select any format over either connection. (Possibly EDID limited options offered by default ?)

2. Source standard - 16-235 for Rec 601 and 709 video stuff, but 0-255 for Windows UI, Photos (and some of the newer non 601/709 digital video stuff)  Ensure that blacks are black and whites are white across content...  However the video is decoded (and de-interlaced) - whether in software or hardware - then surely it isn't too much to ask that broadcast standards are followed consistently. (601 was introduced in 1986 or so - it isn't exactly new. I was working with it in 1990...) 

You have to get 2. right whatever format you are outputting in.  How you ensure 2. is correct, and how sourcea are translated (if required) to the right format selected in 1. is where it gets interesting (and where BTB and WTW could be lost).  There are lots of different approaches (and when I say scale I mean convert between 16-235 and 0-255 not spatially scale) :

a) scale everything to the output format (i.e. scale stuff that is not in the output format, but anything that is leave alone).

b) scale everything to 0-255 internally and then scale again if required on output.

a) could preserve BTB and WTW in 16-235 content, b) presumably couldn't (as <16 and >235 stuff would be scaled to <0 or >255 and thus be crushed/clipped?)

How you scale 8 bit 16-235 dynamic range content to 0-255 is also interesting (in broadcast land there is a Quantel patented Dynamic Rounding technique that dithers to reduce the banding that otherwise might occur in these conversion situations, rather than using a basic static LUT - that will cause issues)  or vice versa.

 

Holy Crap! best thread ever. I have been struggling with terrible live tv performance since I set up my new x64 system. I changed the colorspace setting toYCbCr with a range of 16-235 in Nvidia Control (driver 180) and the difference is incredible.

David, Slider, et al. you guys rule. Thanks much.

---

"There are only 10 types of people in the
world: Those who understand binary, and those who don't"

My VMC HTCP - Vista Ultimate x64
ASUS P5Q, Core 2 Duo 3.15 Ghz
4GB DDR2-800, XFX 9800GT
1 X TB,ATI 550Pro, HDHOMERUN
Linksys 2100, 360, WHS

hamiltonguy:

David, Slider, et al. you guys rule. Thanks much.

Gee thanks

---

-Dave

MCP, MCSA, MCSE 2003
Windows Vista Connected Exp:Home Theater for Technologists
Windows Vista Connected Exp:Home Theater for Sales professionals

Home theater specialist (10+ years)