View Full Version : ATI Radeon HD 2X00 (2400,2600,2900) series owners thread
leeperry
05-16-08, 04:03 PM
didn't really follow, but Seb.26 added some code in fffdshow to actually show on screen the available BTB and WTW dynamically.
it's in the levels section, try Seb's BTB/WTW, and then you got 3 settings, to choose how fast it will react.
then you output in 0-255 RGB32HQ, and you're good to go...and you actually see all the available infos.
HT Slider
05-16-08, 07:23 PM
First of all graphics cards do *NOT* calculate in 32bit the way you believe they do. When a graphics card goes into 32bit mode, that usually means 8bit per color plus 8bit for alpha channel (for half transparency etc). So the data for RGB color information is really only 8bit per color.
But yes, graphics cards usually to convert everything to RGB 32bit sooner or later, but as explained above color information is only 24bit. So it's 4:4:4 24bit.
I was speaking to one of the senior techs at ATI several months ago and was told that all of their video calculations are done using more than 8-bits per color and I'm fairly sure he said 10-bit effectively per color, worst case. He also said there is a lot of "upgrading" going on to the video bandwidth overall.
Even if you are correct and that only 24-bits are used for the calculations, it is still expanded into a higher bandwidth stream (16-235 -> 0-255), even without the 4:2:0 or 4:2:2 to 4:4:4 expansion (which greatly increases the bandwidth).
Displays should be calibrated so that WTW and BTB are not displayed. However, as I said before, some calibrators calibrate the displays so that the first step of WTW and BTB is still visible (and different from pure black/white). Furthermore, I've read that sometimes movies were intentionally authored with deeper BTB and WTW information to make colors look "right". Sounds strange, but that's what I read. Anyway, I just don't want the graphics card to do any violence to my precious video data. If it cuts off BTB and WTW I'm afraid what else will go wrong...
If you read the summary specs on YCbCr over on wikipedia (http://en.wikipedia.org/wiki/YCbCr), all of the color content is supposed to be between 16-235 (or 16-240 for BT.601 for luma only). Nothing except for timing signals are supposed to be outside this range. If it is, it is out of spec.
I have also seen numerous documents that state 16-235 is the required range for BT.709 such as this one http://en.wikipedia.org/wiki/Rec._709
24bit (3 * 8) is a very odd number for a computer. It's like 17.359 for us humans. 32bit (4 * 8) is much more even. It's like 20.000 for us humans. A 32bit CPU can process 32bit data much faster and easier than 24bit data. All the CPU registers are 32bit. All the CPU instructions are optimized for 32bit data. 32bit is just much easier to work with. BUT - 8 of those 32 bits are usually used for the alpha channel and NOT for color information.
I understand that a PC can work with 32-bit more efficiently.
What is "alpha channel" if it doesn't contain any video information?
Can you understand that YCbCr and RGB are not a simple multiply of each other? I don't know the exact formulas for converting one to the other. But in the end there are some funny calculations going on. If you go from 8bit YCbCr 4:2:0 to 8bit RGB rounding has to be done. If you go back from 8bit RGB to 8bit YCbCr, again rounding has to be done. ANY such conversion costs image quality.
The formulas are available here:
http://en.wikipedia.org/wiki/YCbCr
Obviously there will be some rounding going on in the conversion, but if you strictly expand 16-235 to 0-255 rounding the results to integers and then compress back from 0-255 to 16-235 rounding again to integers, you will return to exactly the same numbers. In other words there would be no rounding error at all in the final 16-235 after the expansion and compression. Think of it this way - for every value in the original 16-235 there is only one unique expanded value mapped in the 0-255 range. Going the other way, those mappings will mathematically reverse without any error.
If you run the numbers for the true conversion between 8-bit YCbCr 4:2:2 BT.709 to 8-bit RGB 4:4:4, again you will most likely find a single and unique mapping for every YCbCr value into RGB. Again, this should be reversible without any loss. For certain, you will not be able to "visually" distinguish any error here with normal video. One way to look at this is you are actually mapping lower bandwidth data into a much higher bandwidth and then reversing this calculation. There is plenty of bandwidth in RGB 4:4:4 to provide unique mappings from YCbCr 4:2:2 so there shouldn't be any loss at all if done well.
On the other hand if you were to take original 0-255 content (PC), compress it to 16-235 and then expand that back to 0-255 you would get a round off error here and there. Luckily we are not starting from higher bandwidth PC colors so there is no loss (except for WTW and BTB and this isn't used except for calibration purposes).
First of all a source device shouldn't do *ANY* processing unless it's actually asked to. So ideally a YCbCr 4:2:0 source should just be decoded and sent to the display as it is. Unfortunately HDMI doesn't support YCbCr 4:2:0 transport. But that's not much of a problem. A conversion to YCbCr 4:2:2 is not very complicated and can be done right easily. I believe the hardware decoding chips from Sigma Designs and Broadcom etc output YCbCr 4:2:2, anyway. So ideally the source device should send YCbCr 4:2:2 to the display.
Here I totally disagree with you. If you take high quality, uncompressed video and then using lossy compression, highly compress it into mpeg-2 4:2:0 only to uncompress it back to raw video without any processing, you will end up with literally 1000's of visible artifacts on every single frame.
Instead what is done is the video is always filtered and processed to remove mpeg-2 compression artifacts. This process greatly improves the image quality.
Doing all of this video processing using higher bandwidth RGB 4:4:4 over lower bandwidth YCbCr 4:2:0 should only improve the image quality (if done well).
Now if the source is asked to do some processing, obviously it's an advantage to do it in higher than 4:2:2. So let's go to 8bit YCbCr 4:4:4. That'd still be better than to convert YCbCr to 8bit RGB.
Unfortunately graphics cards usually do everything in RGB. So ok, let the graphics card convert everything to RGB. But still we should try to go as low on processing as possible. That means: Convert YCbCr to 8bit RGB - but without expanding colors to PC levels, please! Because if you convert colors to PC levels and later back to video levels, again you have to round the data two times. This is very likely to introduce banding artifacts.
My feeling is it is best to expand, at least to 4:4:4 as well as perform filtering and processing to remove compression artifacts.
Expanding it further to sRGB (0-255) makes a lot of sense when you consider that we are talking about PCs here and more importantly the latest HDTVs can handle and operate flawlessly with higher bandwidth sRGB vs YCbCr. If the HDTV or monitor can handle sRGB 4:4:4 then there is no reason at all not to "upconvert" to it.
Much more importantly than YCbCr 4:2:0 to sRGB 4:4:4 expansion and potential re-compression (whenever an HDTV not supporting sRGB 4:4:4 is used) is both dealing with all video sources (inputs) and renderers (outputs) using consistent colors as well as having the ability to output in a format that is ideal for the particular monitor/HDTV used. Right now both ATI and Nvidia fail to do these critical jobs in several different ways. At least with ATI we can resort to registry "hacks" to work around most of this, but with Nvidia cards (other than 9000 series) colors and grey levels are inconsistent between different renderers so it is an even bigger mess.
one_2go
05-16-08, 07:47 PM
While I greatly appreciate the education by the post discussing GFX card details, I do believe the recent exchange is a bit Off Topic in the ATI HD2*00 owners thread.
Thanks
While I greatly appreciate the education by the post discussing GFX card details, I do believe the recent exchange is a bit Off Topic in the ATI HD2*00 owners thread.
ThanksThe discussion started out to explain the very significant effect the ATI dongle has on the signal output by the 2xxx and 3xxx series cards. For some time there was a lot of confusion over why some seemed to get video levels from their card and others PC levels and why some reports were of YCbCr and others could get nothing but RGB. The reasons for these differences were not understood until a few weeks ago. It then expanded into a discussion on the merits or otherwise of this process. Both of these I think have a place in this thread. I do tend to agree though that the discussion has run its course.
Hi people, I'm having two annoying problems that I hope you can help with.
I have a 2600 Pro, when I go in ccc, I'm unable to adjust the colors, denois, edge enhancement etc in AVIVO settings. Have you guys seen this problem before?
The other issue is when I watch dvd's or any SD, the red is extremely pixalated! I think I may have read that I need to uncheck the 3:2 pulldown detection but I can't because of the first issue. Any help?
Hi people, I'm having two annoying problems that I hope you can help with.
I have a 2600 Pro, when I go in ccc, I'm unable to adjust the colors, denois, edge enhancement etc in AVIVO settings. Have you guys seen this problem before?
The other issue is when I watch dvd's or any SD, the red is extremely pixalated! I think I may have read that I need to uncheck the 3:2 pulldown detection but I can't because of the first issue. Any help?
Uncheck "use application settings" and you will be able to set them yourself.
It's not checked. It's weird, everything is greyed out.
It's not checked. It's weird, everything is greyed out.
Weird! :confused:
dufflover
05-16-08, 10:36 PM
The other issue is when I watch dvd's or any SD, the red is extremely pixalated! I think I may have read that I need to uncheck the 3:2 pulldown detection but I can't because of the first issue. Any help?
Is it something like what I had before?
http://www.avsforum.com/avs-vb/showthread.php?p=13708327#post13708327
I found my problem! It turns out that it was simply because I had it hooked up on the second DVI port! Once I switched it to the first port, the options worked fine.
I was speaking to one of the senior techs at ATI several months ago and was told that all of their video calculations are done using more than 8-bits per color and I'm fairly sure he said 10-bit effectively per color, worst case.
I can tell you as a fact that there is a specific list of video formats supported between different DirectShow filters and the predefined RGB32 format contains 24bit of color information and 8bit alpha channel. The alpha channel is used for transparency and is virtually useless for movie playback.
If you run the numbers for the true conversion between 8-bit YCbCr 4:2:2 BT.709 to 8-bit RGB 4:4:4, again you will most likely find a single and unique mapping for every YCbCr value into RGB. Again, this should be reversible without any loss.
I find it funny how you construct your own guesses and ignore what experts are saying. Do you think you know better than the technical director of Broadcom? (see link in my last post)
If you take high quality, uncompressed video and then using lossy compression, highly compress it into mpeg-2 4:2:0 only to uncompress it back to raw video without any processing, you will end up with literally 1000's of visible artifacts on every single frame.
Instead what is done is the video is always filtered and processed to remove mpeg-2 compression artifacts. This process greatly improves the image quality.
I'm sorry, but it seems to me that you have no clue about how these things work. A video decoder always ends up with YCbCr data after decoding is completed with all mandatory steps (which includes deblocking for h264). There are additional compression artifact reduction algorithms available like e.g. mosquito noise reduction and block artifact noise reduction. These algorithms are at this time *NOT* available by any graphics card. Furthermore experts strongly suggest to only enable these algorithms if they are really needed cause they help only with really bad sources while they do damage to good sources. Finally, these algorithms are usually done in YCbCr and not in RGB. Also deinterlacing and scaling is done by video processors in YCbCr and not in RGB. The only thing a good video processor (e.g. Lumagen Radiance) does in RGB is color correction. Everything else is done in YCbCr.
This will be my last post about this topic in this thread. Sorry for going slightly OT, guys.
This will be my last post about this topic in this thread. Sorry for going slightly OT, guys.
Thank you for going off topic. I got a lot of valuable information out of it which makes it even more easy to track down certain problems because you get an idea what's going on under the hood.
grubi.
I apologise for baiting the OT exchange. I had a feeling a few people might object to the over-simplification of the exchange in posts #5223 & 4 but I didn't realise the topic was still quite as emotive and full of different opinions.
Having said that, like grubi, I thoroughly enjoyed reading the comments and many will appreciate the information supplied (even if some posts contradict others) ;)
On Topic (mostly), I'd completely missed the discovery that the ATi dongle forced the card to convert it's ouput to YCbCr. This explains a lot. Thanks.
Wo0zy
Which driver set is the latest, VisionTek Catalyst 8.3 or ATI Catalyst CC 8.4. I'm running Vista Ult. with a HD2600 XT and is slowly coming up to speed on playing HD material and the HA issues.
Can someone point me to the summary page with the tweaks and solutions so far?
Thanks
arfster
05-17-08, 01:03 PM
Allo all - just been experimenting some with the 8.4 denoise/detail settings (been meaning to do this for ages).
Detail: as before, this only seems to affect SD (at least on my 2600xt, but I expect newer cards will be the same, as this has a minimal hit on shaders). As someone else pointed out, sharpening at 50 seems to be a good compromise for SD - perhaps even anything less is a negative value? I checked the dlls, and there's no reg setting to force HD sharpening.
Denoise: does work with HD (maybe not on 24xx/34xx models?), but only with interlaced stuff. For SD it's OK, but the default setting 64 is a bit much. Around 30ish seems better - gets rid of irritating dark scene noise and wipes quant noise without much of a hit on detail. For HD it wipes out too much detail even at 30, so I used vforcehddenoise=0 to switch it off.
Basically, with this setup you get only an effect on SD. The exact levels you want, particularly of denoise, depend largely on the quality of content you're watching. If you're upscaling 3mbit mpeg2 broadcast channels, more is probably better :-)
bobbyrr
05-17-08, 09:37 PM
Wow. Now that is interesting. Are you able to test BD playback?
I thought ATi always expanded BD playback to PC levels. If the adapter stops this from happening what exactly do we end up with?
Wo0zy
BD colours look normal to me.
Actually I have another question related to BD, I try to play it in powerdvd 8 ultra, however, the video playback is not smooth, but rather jerky.
I have AVIVO installed, and the playback preformance is the same (jerky) with either hardware acceleration ticked or unticked.
HT Slider
05-17-08, 11:54 PM
I can tell you as a fact that there is a specific list of video formats supported between different DirectShow filters and the predefined RGB32 format contains 24bit of color information and 8bit alpha channel. The alpha channel is used for transparency and is virtually useless for movie playback.
A quick Google search demonstrates that ATI's own specification for ATI HD AVIVO includes: "Full 30 bit display processing" for all of their HD cards as well as supporting 30 bit displays? I find it difficult to believe they are talking about transparency when it comes to AVIVO, HD and video processing.
http://ati.amd.com/products/Radeonhd2600/specs.html
Either you or ATI is incorrect.
Also, as I mentioned before, I was personally told by a senior tech at ATI that they use 10 bits per color during their image processing (I am certain now that it was 10 bits). Note that ATI's specification states 30 bits and 3 colors X 10 bits = 30 bits.
I find it funny how you construct your own guesses and ignore what experts are saying. Do you think you know better than the technical director of Broadcom? (see link in my last post)
I find it funny that you feel it necessary to turn this into a bashing session, instead of openly discussing the facts. You have not presented a single piece of evidence that ATI's transform functions are flawed. Mathematically there is no requirement that there be any image degradation due to these transforms.
I did read your old link to a reference from the Broadcom Director, but I don't see how that proves anything to do with hardware and software that wasn't even available to the director at the time.
While I can't say for certain that ATI doesn't produce errors in their transform functions, I do know what I have learnt during 15 years working in satellite and military video and night vision imaging systems. I can tell you with absolute certainty that when you take a low bandwidth stream of video and upconvert or encode that data into a higher bandwidth stream of video and then finally compress it back into the original low bandwidth stream that there is absolutely no reason for there to be any loss of accuracy. There is also no reason for there to be any additional error going from a low bandwidth to high bandwidth followed by a medium bandwidth vs simply going from the low straight to the medium. The mathematicians I've worked with have never created video transforms that do. Also, looking at the simple linear equations used to transform between YCbCr and sRGB I am absolutely certain that it would be relatively easy to create transform equations that are not lossy and that can be reversed without any loss of accuracy. Even if you were to perform the calculations using brute force, without any regard for accuracy, it would be very unlikely that there would be any detectable error when watching actual video.
Considering there is discrepancy between your stated 24 bit video processing and ATI's stated 30 bit processing, it is still possible that 30 bits are used and this would make it extremely difficult for any transform errors to occur - even using the simplest, brute force calculations. If you notice, ATI's spec also states that their HD cards support 30 bit filtering, 30 bit dithering, and even 30 bit color displays (all within the AVIVO portion of the spec) - suggesting that the entire pipeline supports 30 bits; only to be downconverted to 24 bits for our final transform from RGB to YCbCr so our lower bandwidth HDTVs can display our movies.
We have also had other members on this forum state that they can't detect any difference between viewing the content in sRGB compared to being reconverted back to YCbCr.
Do you have any actual proof that the video quality is degraded through this process? Do you have any links where someone has taken a digital YCbCr 4:2:0 BT.709 stream, sent it through an ATI HD card and then examined the end result to look for specific errors that can be attributed to the YCbCr 4:2:0 -> RGB 4:4:4 -> YCbCr 4:2:2 conversion?
Keep in mind the bandwidth ratios through the various transforms are effectively 6:14:8; meaning there is much more bandwidth available at the middle (RGB) stage. Mathematically there is no reason for any numerical difference between that and simply going from YCbCr 4:2:0 -> YCbCr 4:2:2 (6:8 bandwidth change). Note that if ATI is indeed using a 10 bit RGB 4:4:4 stage for processing, then the bandwidth ratios become 6:56:8.
I'm sorry, but it seems to me that you have no clue about how these things work. A video decoder always ends up with YCbCr data after decoding is completed with all mandatory steps (which includes deblocking for h264). There are additional compression artifact reduction algorithms available like e.g. mosquito noise reduction and block artifact noise reduction. These algorithms are at this time *NOT* available by any graphics card. Furthermore experts strongly suggest to only enable these algorithms if they are really needed cause they help only with really bad sources while they do damage to good sources. Finally, these algorithms are usually done in YCbCr and not in RGB. Also deinterlacing and scaling is done by video processors in YCbCr and not in RGB. The only thing a good video processor (e.g. Lumagen Radiance) does in RGB is color correction. Everything else is done in YCbCr.
This will be my last post about this topic in this thread. Sorry for going slightly OT, guys.
Please - don't try to be so belittling and don't be ridiculous.
The video cards that this thread is about all do very significant amounts of noise reduction, by default. Deblocking itself is a noise filter specifically applied to remove compression artifacts and it is most definitely used by our ATI cards. To argue that no denoise algorithms are used to remove compression artifacts is simply wrong. They may not be using "mosquito noise reduction and block artifact noise reduction", but they are most certainly using noise reduction that helps to hide compression artifacts.
In addition to deblocking, evidence of compression artifact reduction through additional filtering can easily be visualized by performing the following experiment:
Take digitally recorded standard definition content that is moderately compressed (relatively low bitrate - such as typical of a Satellite SD broadcast) and play it using an ATI HD 2600XT. Using default settings, it will produce an image with greatly reduced compression artifacts compared to if you play it back after adding the TRDenoise=0 registry setting (add and remove the registry entry while stopping and playing the same scene). I have done this myself - just a few days ago. The image looks much better with the temporal denoise algorithms in place. With TRDenoise=0 applied you will see 1000's (and I do mean 1000's) of visible compression artifacts in the form of blocks.
HT Slider
05-18-08, 03:29 AM
I've been doing some more research into this and according to my contacts (I don't want to drag any of them into this) and further reading:
Once the compressed video stream is sent to the video card for DXVA processing, the uncompressed video is never sent back to the CPU. iDCT, motion compensation, de-blocking, subsampling format conversion, color space conversion, color correction, deinterlacing, inverse telecine, noise reduction filtering, detail enhancement, scaling, etc. and ultimately rendering are all performed directly by the video card.
Video cards with AVIVO HD perform all of their video processing using 10 bits per color. See the following link and the section on "True-to-life image reproduction"
http://ati.amd.com/technology/Avivo/features.html
The HD 2400 series (also an AVIVO HD card) processes all video using 10 bits per color too (30 bits per pixel):
http://ati.amd.com/products/radeonhd2400/specs.html
When the output is downconverted for a 24-bit monitor or HDTV hooked up through DVI, RGB 4:4:4 is used. Dithering is also used to increase the effective number of colors from 24-bit to 30-bit.
When the output is sent to an HDTV through HDMI, the 30 bits per pixel are typically output at 24-bit YCbCr 4:4:4.
Assuming there actually is an RGB to YCbCr conversion at #4; Due to the 30 bit processing and the far greater bandwidth available during the 30 bit RGB 4:4:4 processing stage, no round off errors would occur in the double conversion.
Another interesting find is that DirectShow does actually support YCbCr and directshow filters can operate in this mode as opposed to RGB. Also, the VMR7 and VMR9 renderers both support YCbCr directly. Here are some links worth looking at for those interested:
http://msdn.microsoft.com/en-us/library/ms867704.aspx
http://msdn.microsoft.com/en-us/library/ms788177.aspx
http://msdn.microsoft.com/en-us/library/ms788178(VS.85).aspx
One interesting quote from the following page:
http://msdn.microsoft.com/en-us/library/bb530104(VS.85).aspx
"Also, the conversion from RGB to YUV is not lossy"
Here are a couple of links that discuss standard HDMI and DVI video streams (YCbCr vs RGB) and potentially explain why ATI chose to use YCbCr when their dongle is used and RGB when DVI is used (or a non-ATI HDMI dongle):
http://www.hdtvmagazine.com/forum/viewtopic.php?t=5317
http://www.hdtvmagazine.com/forum/viewtopic.php?t=5427
Also note that there are still quite a few unknowns, including:
How do we know for certain that the colorspace is actually converted from YCbCr to RGB when the output being used is YCbCr. DirectShow supports YCbCr, and I can't think of any reason ATI couldn't choose to do all of their video processing in YCbCr if they wanted to. Using YCbCr throughout would save quite a bit of processing load, especially since it doesn't normally pass video back to DirectShow. Screen captures can easily be converted to RGB on a single frame basis.
I've been doing some more research into this and according to my contacts (I don't want to drag any of them into this) and further reading:
Once the compressed video stream is sent to the video card for DXVA processing, the uncompressed video is never sent back to the CPU. iDCT, motion compensation, de-blocking, subsampling format conversion, color space conversion, color correction, deinterlacing, inverse telecine, noise reduction filtering, detail enhancement, scaling, etc. and ultimately rendering are all performed directly by the video card.
Video cards with AVIVO HD perform all of their video processing using 10 bits per color. See the following link and the section on "True-to-life image reproduction"
http://ati.amd.com/technology/Avivo/features.html
The HD 2400 series (also an AVIVO HD card) processes all video using 10 bits per color too (30 bits per pixel):
http://ati.amd.com/products/radeonhd2400/specs.html
When the output is downconverted for a 24-bit monitor or HDTV hooked up through DVI, RGB 4:4:4 is used. Dithering is also used to increase the effective number of colors from 24-bit to 30-bit.
When the output is sent to an HDTV through HDMI, the 30 bits per pixel are typically output at 24-bit YCbCr 4:4:4.
Assuming there actually is an RGB to PCbCr conversion at #4; Due to the 30 bit processing and the far greater bandwidth available during the 30 bit RGB 4:4:4 processing stage, no round off errors would occur in the double conversion.
Another interesting find is that DirectShow does actually support YCbCr and directshow filters can operate in this mode as opposed to RGB. Also, the VMR7 and VMR9 renderers both support YCbCr directly. Here are some links worth looking at for those interested:
http://msdn.microsoft.com/en-us/library/ms867704.aspx
http://msdn.microsoft.com/en-us/library/ms788177.aspx
http://msdn.microsoft.com/en-us/library/ms788178(VS.85).aspx
One interesting quote from the following page:
http://msdn.microsoft.com/en-us/library/bb530104(VS.85).aspx
"Also, the conversion from RGB to YUV is not lossy"
Also note that there are still quite a few unknowns, including:
How do we know for certain that the colorspace is actually converted from YCbCr to RGB when the output being used is YCbCr. DirectShow supports YCbCr, and I can't think of any reason ATI couldn't choose to do all of their video processing in YCbCr if they wanted to. Using YCbCr throughout would save quite a bit of processing load, especially since it doesn't normally pass video back to DirectShow. Screen captures can easily be converted to RGB on a single frame basis.
That all seems about right and pretty much what I said in #5222......... but it seems internal 10-bit processing stops this from causing image degradation.
Regarding how we are sure a double conversion takes place, well we certainly know that if you screengrab an image from HD video video black is "0" not "16", even though it is "16" when it arrives at an HDTV connected using the dongle. I know these are levels rather than colorspace, but it is an indication of conversion going on. The video is clearly expanded losing all blacker than black and whiter than white.
And unless you are using full screen exclusive mode, which PowerDVD/TMT and other current Blu-ray players do not use, Windows needs to be able to mix video with desktop elements, like other applications/start menu/taskbar etc. It is really hard to see how that would work without all sharing the RGB colorspace. As we said earlier, I can imagine it being possible to write a fullscreen custom renderer that works entirely in YUV, but i see no evidence it has been done.
You are right we do not have all the facts but personally I am sure that everything is first converted to RGB 0-255 and then back to YCbCr 16-235.
@HT Slider, if you want to discuss this further you should open a new thread (and PM me the link). I don't think it makes sense to do a full discussion of this topic in this thread. We would only annoy people.
tetsuo55
05-18-08, 05:01 AM
I have my videocard hooked up to my tv with a DVI to HDMI cable.
I Basically want everything to be TV levels, as the tv is obviously only capable of displaying those, what settings do i need to change to turn everything into TV levels and to make sure not double conversion is taking place.
@HT Slider, if you want to discuss this further you should open a new thread (and PM me the link). I don't think it makes sense to do a full discussion of this topic in this thread. We would only annoy people.
If you continue discussing at a different place, please be so kind to PM me the link also.
Thanks a lot.
grubi.
Which driver set is the latest, VisionTek Catalyst 8.3 or ATI Catalyst CC 8.4. I'm running Vista Ult. with a HD2600 XT and is slowly coming up to speed on playing HD material and the HA issues.
Can someone point me to the summary page with the tweaks and solutions so far?
Thanks
Anyone?
I have connected my TV (Pana TH50PZ700) via the ATI DVI/HDMI dongle to a HD2600XT. Due to the discussion here I tried with a 3rd party HMDI/DVI cable, however I can't see any difference. How can it be verified if the card is outputting RGB or YCbCr? Is it possible that display and graphic card negotiate in which mode to operate and YCbCr is choosen as prefered regardless if the ATI dongle is used or not? Does anybody have some insight knowledge if such negotiation is part of the DVI/HDMI protocol spec?
grubi.
HT Slider
05-18-08, 01:23 PM
I have connected my TV (Pana TH50PZ700) via the ATI DVI/HDMI dongle to a HD2600XT. Due to the discussion here I tried with a 3rd party HMDI/DVI cable, however I can't see any difference. How can it be verified if the card is outputting RGB or YCbCr? Is it possible that display and graphic card negotiate in which mode to operate and YCbCr is choosen as prefered regardless if the ATI dongle is used or not? Does anybody have some insight knowledge if such negotiation is part of the DVI/HDMI protocol spec?
grubi.
I just picked up an HDMI/DVI cable also to see if I can convince my 2600XT to output using YCbCr and TV/video grey levels for my HDTV but haven't had a chance to try it yet. In my case I'll be able to see if PC or TV/video grey levels are used because my HDTV definitely only supports 16-235 and it definitely doesn't negotiate this when DVI -> DVI is used (without an ATI HDMI dongle; I'm going to try the dongle combined with the HDMI/DVI cable).
I don't know any way to verify for certain if YCbCr or RGB is used, nor if RGB 0-255 or RGB 16-235 is used (or if our cards even support RGB 16-235).
One thing for certain is the HDMI and DVI specs do include negotiation. The display and source are supposed to negotiate appropriate communication protocols to use. Based on what I've read, when HDMI is used YCbCr and TV/video grey levels are supposed to be favored. When DVI is used, RGB and PC grey levels are supposed to be favored. In either case, negotiations can force one over the over. The deciding feature for favoring one over the other for our cards appears to be if an ATI HDMI dongle is used.
Also, there are HDTVs that do properly support RGB with PC levels and yours might be one of these. The only thing I can think of is you might see slightly different amounts of banding (possibly improved) when RGB 0-255 is used over YCbCr.
I should have a chance to test mine out this evening (or afternoon; if it rains) and I'll post my results.
Ultimately we need to push ATI to include a 4-way selector for output formats when an HDTV is used:
Automatic
RGB 0-255
RGB 16-235
YCbCr (16-235)
I don't know how well they listen, but suggestions can be sent directly to ATI's "Catalyst Crew" (driver development team) through this link:
http://support.ati.com/ics/survey/survey.asp?deptID=894&surveyID=486&type=web
HT Slider
05-18-08, 02:27 PM
Anyone?
arfster is the guy who figured out and published all of the registry settings: http://www.avsforum.com/avs-vb/showpost.php?p=11659897&postcount=2121. In the past (previous driver releases), these registry tweaks were fundamental for using these cards for HTPC use.
If you are running the latest driver (Catalyst 8.4) with a 2600XT and a single display, the only things you probably want to consider for the registry are:
"TrDenoise"="0"
and
"VForceMaxResSize"="2800000" (and only if you are using a very high resolution display; my HDTV running at 1920x1080 using 1080i does not need this, so I assume only higher resolutions need this)
The TrDenoise registry entry is only required when your source video has already been highly "denoised". With digital content, the image quality is considerably better (IMO) without this entry. On the other hand, if you capture through s-video, your STB does a lot of denoise and the TV Tuner used to capture does a lot more, then you probably will need to add TrDenoise=0 in order to ensure you don't get "ghosts" in the image with certain shows (films recorded over cable/satellite in SD are the worst for this in my experience).
I actually don't use any registry tweaks with our setup anymore (but with previous drivers versions needed to use many).
You probably will want to make some adjustments within the Catalyst Control Center though - dependent on the type of display used:
If your HDTV is hooked up using DVI, you might need to change the overall brightness to +15 and contrast to 85% to see appropriate grey levels (without this the image may be too dark in dark scenes and too bright/clipped in bright scenes or bright objects). For an explanation as to why this may be required, read up on YCbCr vs RGB in this thread (essentially many HDTVs only support TV/video grey levels (YCbCr), but the video card outputs using PC grey levels (RGB) through DVI).
If you are using Vista and Media Center, you can use C:\Windows\ehome\Brightness.wmv to help adjust the overall brightness slider (so you can "just" see the moving X) and C:\Windows\ehome\Contrast.wmv to adjust the overall contrast (so the shirt shows fine detail and doesn't look like a solid white). Since we use Media Center regularly, I prefer to make adjustments using Media Center but with Catalyst 8.4 there is no difference at all in grey levels (on my system at least) between Media Center and Media Player so it shouldn't really matter.
If you really want to optimize everything, first crank up the brightness and turn down the contrast on your HDTV; then use the Brightness.wmv and Contrast.wmv files to adjust the video card's output to match your HDTV (you'll find either the default 0/100% or +15%/85% (or possibly +16/86%) are the only options that could be required; depending on your HDTV and the hookup). After that, adjust your HDTV itself so the moving X is just visible and the shirt isn't excessively bright. Then as a final step, if you really want to go all out, use your preferred calibration DVD/Blue-ray to test and tweak your calibration.
If your HDTV is hooked up using HDMI and it requires PC grey levels, you might need to replace the ATI HDMI dongle with a generic HDMI/DVI adapter (without this the image is washed out). This issue isn't very common and only certain HDTVs bring out this issue. The source of the problem is the opposite to #1 above.
Depending on your HDTV, you may prefer to turn down the denoise slider in CCC. In my experience 50% is better for my rear projection Toshiba, but my fathers Olevia produces a better image for SD with it set to 70%.
Depending on your HDTV, you may prefer more or less edge enhancement. I prefer 15% on my rear projection Toshiba, but my father's Olevia 747i produces a better image for SD with it set to 0% and disabled alltogether.
Turn down and disable both colorvibrance and fleshtone correction (I haven't found anyone who prefers the image with these enabled).
EDIT: My comments on not needing the registry tweaks are only true for 2600 series cards, the latest drivers and systems using PCIe video cards. 2400's, systems using AGP video cards and older drivers do require additional registry tweaks.
RGB vs YCbCr is not easy since apart from any rounding/computational errors they should look identical. To know, you need a device which tells you the signal it is receiving. A couple of people with Lumagen Video Processors, which can do this, have confirmed the change from RGB to YCbCr when the dongle is used.
Confirming 0-255 or 16-235 is relatively easy if you are proficient at calibrating. First calibrate to video levels using a known video level source, then attach your HTPC with default driver settings (no brightness or contrast adjustments). If the levels are the same you have 16-235. If you get crushed blacks and blown whites you have 0-255.
These results have now been verified by several people. One thing may be confusing things. There are two versions of the dongle.One for the 2xxx series one for the 3xxx series.
HT Slider
05-18-08, 11:50 PM
I finally got around to trying the ATI HDMI dongle and my HDMI to DVI cable.
I have to say I'm more confused than ever now...
The instant I installed the cable, I noticed a distinct improvement in color (or at least "more" color - blues and reds are deeper) as well as a desktop where darker objects in the image were much brighter.
So - on the surface it looked like everything was going to work properly and I could resort to using default brightness and contrast levels. Here is where things get weird though...
With my previous DVI to DVI connection (note I have gone back and forth between DVI->DVI and ATI HDMI dongle->HDMI/DVI cable a couple of times tonight to try to confirm what I am seeing):
My HDTV suffered from the expected problems when the display uses video levels (16-235) and it is being driven with PC levels (0-255).
To fix this, in CCC I had to crank up the overall brightness to +15 and lower the overall contrast to 85%. Once this was done, everything as far as wmv, mpeg-2, dvr-ms, DVD, pictures (still a little too dark actually), desktop and video games all displayed using correct grey levels (using Media Center or Media Player for video content). This also makes sense when you consider I was faking the output to make it comply with video levels (16-235).
In order to get PowerDVD Ultra to play correctly though I found I had to set the AVIVO brightness and contrast to +15/85% and this caused Media Center & Media Player to over compress the colors (due to the combined overall +15/85% and the AVIVO +15/85%).
In order to get both PowerDVD Ultra and video content within Media Center/Media Player to play correctly I found I had to actually use only the AVIVO +15/85%, but this breaks the ability to display photographs, desktop & video games correctly.
With the new connection using the ATI HDMI dongle and the HDMI->DVI cable to my HDTV:
Using totally default settings, the first thing I noticed with my desktop is the colors are deeper and brighter.
Using my AVIA calibration DVD, played with Media Center, shows that my system is perfectly calibrated for DVD playback (as far as brightness and contrast is concerned) using default CCC settings. At this point I was very pleased!...
Playing Microsoft's wmv test clips in Media Player or Media Center, my brightness is correct (at 0/default), but my contrast needs to be dropped down to 85% to prevent clipping of the whites. This can be either in the overall settings or in AVIVO. Retesting the AVIA DVD with the new setting, it now shows that the contrast is too low (and once again, bringing the contrast back to the default 100% is correct).
Playing an HD dvr-ms recording in Media Center, it too shows that contrast levels are way to high and I need to use a contrast of 85% to prevent clipping with certain shows (in particular CBC news in 1080i and the news announcer's white shirt). Still, a brightness of 0 (default) is still correct (trying +15 for brightness causes black objects to be a grey and not black)!
If I take a Blueray movie (Harry Potter and the Order of the Phoenix in this case) and play it using PowerDVD Ultra, it once again suffers from the entire PC level issue. To make it display correctly I have to set both the brightness to +15 and the contrast to 85% again and these have to both be set within the AVIVO panel (PowerDVD somehow ignores the overall brightness and contrast settings).
If I take my test "photograph" ranging from 0-255 grey levels, I have to configure the overall brightness and contrast to +15/85% to see the full 0-255 grey levels without clipping.
Essentially everything is totally messed up whether I use or don't use the HDMI dongle. In the past I thought everything was working (without the dongle), except for the "glowing" issue in PowerDVD, but tonight I discovered that this was actually because PowerDVD requires the AVIVO settings configured for +15/85% no matter if I use the ATI HDMI dongle or not.
Without the HDMI dongle I can get either everything to do with video working correctly, or Media Center combined with pictures, desktop and games; but not both.
With the HDMI dongle I haven't found a way to get PowerDVD to produce correct grey levels without messing up Media Center.
Does anyone have their Vista Media Center PC, with an ATI HD video card and an HDTV, truly producing correct grey levels accross all media (pictures, wmv, dvr-ms, DVD, Blue-ray, HD-DVD, etc.)?
Does anyone have their Vista Media Center PC, with an ATI HD video card and an HDTV, truly producing correct grey levels accross all media (pictures, wmv, dvr-ms, DVD, Blue-ray, HD-DVD, etc.)?
No! Even with the UseBT601CSC registry tweak I get all sorts of grey levels with diff applications and diff media.
Really starting to piss me off.
arfster
05-19-08, 03:52 AM
HT Slider: if possible, could you do an ATI bug report with your post above? Sums up the absurdities of forced video expansion.
ps a sarcastic comment about the need for usebt601csc=1 registry tweak to get SD levels the same woudln't go amiss either :-)
>>Using my AVIA calibration DVD, played with Media Center, shows that my
>>system is perfectly calibrated for DVD playback (as far as brightness and
>>contrast is concerned) using default CCC settings. At this point I was very
>>pleased!...
This is the same with my setup however I use MediaPortal for playback. I also have to apply the UseBT601CSC registry switch.
>>Playing Microsoft's wmv test clips in Media Player or Media Center, my
>>brightness is correct (at 0/default), but my contrast needs to be dropped
>>down to 85% to prevent clipping of the whites. This can be either in the
>>overall settings or in AVIVO. Retesting the AVIA DVD with the new setting, it
>>now shows that the contrast is too low (and once again, bringing the
>>contrast back to the default 100% is correct).
Can you post links for the testfiles then I will be able to test it on my system.
grubi.
Because of the problems reported here I have done a little more testing myself and it adds yet another twist to this convoluted tale!
On my XP system all is good -with the dongle and UseBT601 tweak all definitely correctly calibrated with default settings in CCC - pictures, SD video and HD video, uisng Directshow players and PowerDVD7 or 8 (SD and HD).
In testing I found one new thing. All my previous testing had been using VMR9. If I use OVERLAY I get perfect levels but without clipping BTB and WTW. It seems that overlay is able to pass 16-235 levels strgiht through without double conversion. I do not have the equipment to say if it is passing YCbCr or RGB, but my guess would be YCbCr. I will try to get this checked by someone with a Lumagen.
NB: I only tested this with SD (DVE). I do not have an HD test pattern with BTB. If anyone has one they can link to then I will test that too.
arfster
05-19-08, 08:38 AM
NB: I only tested this with SD (DVE). I do not have an HD test pattern with BTB. If anyone has one they can link to then I will test that too.
There's one at http://w6rz.net/ , the 1-255 ramp.
As an aside, in XP it's possible to break expansion a few ways (never tested overlay though), but the colours generally get screwed up instead. If you measure the values from the 75% smpte pattern at the above site it's really obvious when this happens.
I'll give it a go. Thanks. Interesting though that the levels were right without expansion leading to clipping.
You are right, colors are wrong :(. How many different ways can a driver behave? They must have worked pretty hard to mess things up this badly.
You are right, colors are wrong :(. How many different ways can a driver behave? They must have worked pretty hard to mess things up this badly.
Both with VRM9 and overlay?
grubi.
arfster
05-19-08, 10:53 AM
Can't remember if I noticed this before: if you use ffdshow to feed RGB32HQ to EVR then levels aren't expanded. Maybe not that useful to a lot of people because it kills acceleration, but at least it tells us that ATI driver expansion is (probably) happening in the RGB conversion, rather than being some deliberate act of postprocessing. I wonder if it's possible to hack the BIOS and alter the matrix........
Note if you're doing this remember to use different ffdshow profiles for above and below 1024 horizontal resolution to select bt601 and bt709 respectively (cos the ATI auto-select will no longer work, as you're bypassing the card's RGB routines). Also note hardware deinterlacing won't work with this, as it only accepts nv12.
arfster
05-19-08, 11:08 AM
Can't remember if I noticed this before: if you use ffdshow to feed RGB32HQ to EVR then levels aren't expanded.
......
Also note hardware deinterlacing won't work with this, as it only accepts nv12.
Presumably for the same reason, sharpening/denoising are disabled also. The drivers seem to have no processing effect whatsoever on the whole video process when you do the ffdshow yuy>RGB, other than in scaling routines (and you can ffdshow there if you want).
gbcrush
05-19-08, 11:49 AM
HT Slider: if possible, could you do an ATI bug report with your post above? Sums up the absurdities of forced video expansion.
ps a sarcastic comment about the need for usebt601csc=1 registry tweak to get SD levels the same woudln't go amiss either :-)
(Havent logged in for ages, but this very issue was itching at my mind).
Not saying this to bash ATI, but would the "easy fix" to the above solution be going with an nvidia card? A quick peek into the recent "Avivo vs. Pure.." thread makes me think "not quite".
I'm worried about the color levels on my TV though. I'm guessing they're off, as I've set them so that SD MPEG2 through my VMC box looks ok (so TV, DVD, VMC calibration videos look ok). That means everything else like pictures, WMV, any HD source is going to be off, right?
(I think) it gets worse though, since the TV is actually being fed by HDMI via my receiver. So, if I were looking at a standalone Blu-ray player in my future, its video levels would be off too, right? As would any of the game consoles I have feeding into the receiver and then out to the TV via HDMI.
For reference, I haven't done any registry tweaks for the 2600pro I have with regards to the video levels. I had simply calibrated my TV's settings to show decent color for SD mpeg2 playback. Now that I'm growing (and planning on growing) what's being connected to the receiver, I thought maybe I needed to use the registry tweaks to bring MPEG2 video in line, but it looks like there's no 100% solution?
Oh. PowerDVD is not a concern, yet.
Sorry for the rambling :)
Both with VRM9 and overlay?
grubi.No, just with Overlay.
Can't remember if I noticed this before: if you use ffdshow to feed RGB32HQ to EVR then levels aren't expanded. Maybe not that useful to a lot of people because it kills acceleration, but at least it tells us that ATI driver expansion is (probably) happening in the RGB conversion, rather than being some deliberate act of postprocessing. I wonder if it's possible to hack the BIOS and alter the matrix........
Note if you're doing this remember to use different ffdshow profiles for above and below 1024 horizontal resolution to select bt601 and bt709 respectively (cos the ATI auto-select will no longer work, as you're bypassing the card's RGB routines). Also note hardware deinterlacing won't work with this, as it only accepts nv12.That is odd. Expansion still happens with XP/VMR9 and RGB32HQ conversion.
By the way, although you need different profiles for HD and below, you can also specify different output colorspaces for each profile. So in my case I use YV12 for HD and RGB for upscaled SD.
HT Slider
05-19-08, 05:45 PM
No! Even with the UseBT601CSC registry tweak I get all sorts of grey levels with diff applications and diff media.
Really starting to piss me off.
This is annoying, but after lots of messing around I think I have the best compromise for my Toshiba 51H83. BTW, the UseBT601CSC registry tweak doesn't seem to do anything at all since I installed Catalyst 8.4 on my system (I haven't checked since I switched to the HDMI dongle, but it definitely doesn't do anything without the dongle being used).
I discovered that if I leave the AVIVO brightness and contrast at default (configured by application), that I can use PowerDVD Ultra's advanced video configuration to set the AVIVO brightness and contrast. Unfortunately the scales on the sliders do not correspond to the same values within CCC, but they do work (for Blue-ray anyway). Flipping back and forth between using AVIVO settings and PowerDVD Ultra settings I think I have a reasonably close setup.
Here is what I am currently using:
ATI HDMI dongle + HDMI->DVI cable
Overall brightness/contrast in CCC: +4/85%
AVIVO brightness/contrast in CCC: default and controlled by application
Brightness/contrast in PDVDU: +20/-5
With the above settings I get:
Close to correct brightness/contrast in PowerDVD Ultra (note I do not own a Blue-ray calibration DVD and the settings seem to only affect Blue-ray playback)
Playback of wmv, dvr-ms and other video types within either Media Center or Media Player produces very slightly too much brightness (using +4 and it really should be 0). It isn't bright enough to really make blacks turn into grey when watching movies & TV, but I can fairly easily see the moving BTB bars and moving X in various test patterns.
Playback of DVDs within Media Center or Media Player produces very slightly too much brightness and nowhere near enough contrast. BTB test patterns do show moving bars and X's, but it isn't enough to make blacks really turn grey. WTW test patterns show that I am way down on the white end of the grey scale. To pass the tests, I really need to use brightness at 0 and contrast at 100 (default settings). The end result when watching DVDs is movies are a little dull, but still look good.
Playback of Blue-ray movies within PowerDVD Ultra produces what looks like accurate grey levels, but I am only using black scenes and objects to calibrate brightness and very bright white objects to calibrate contrast. If I pick up a calibration Blue-ray I can fine tune PowerDVD to make it spot on (since PowerDVD can take control of the AVIVO settings).
Viewing of photographs, desktop (PC applications), and video games the grey levels are still too dark on the brightness side of things. Dark objects are black and all grey levels below 10 (on the 0-255 scale) are completely black. On the white side, I can see white variations all the way up to 255 without any clipping.
I think this is the best I've actually ever had it (now that I realized out that PowerDVD wasn't displaying correctly with my previous DVI/DVI cabling).
With this setup I have the following problems:
I am missing the bottom 0-10 grey levels (out of 0-255) for photographs, PC applications and video games (if I manually increase the overall brightness to +15 I can see all 0-255, but then brightness is too high for video playback)
I am able to see BTB within the range 10-16 for DVDs, wmv, dvr-ms, and other video formats within both Media Center and Media Player.
DVDs (only) are not making use of full brightness so the whitest whites are not as bright as they should be.
PowerDVD Ultra does not have exactly the correct grey levels, but only because I don't own a Blue-ray calibration DVD.
I have to say this situation really is silly when you consider that Media Center PCs are being sold as high end all around Media display devices - yet grey levels are actually impossible to get correct for different types of media simultaneously. I can get grey levels 100% accurate for any type of media - just not without making adjustments every time I switch between different types of media.
I have borrowed an Nvidia 8500GT and am right now going to install it. Unfortunately it doesn't have an HDMI dongle so I suspect by default it will output using RGB and PC levels. It will be interesting if I can tweak and tune to get accurate grey levels throughout with this video card...
Ultimately I am planning to replace my HD 2600XT since it keeps crashing when playing HD content within Media Center (some sort of DXVA initialization glitch). I'm planning to either go with an ATI 3870 or an Nvidia 8800 GT. Whichever one is able to drive my Toshiba HDTV is the one I will be going with.
Note that not only are grey levels important, but automatically dealing with overscan when all PC resolutions are used and being able to output spec compliant 720p and 1080i/p (complete with overscan) - without requiring me to go into the video card control panel each time. In other words I want to be able to watch TV/Blue-ray/HD-DVD/etc. with the best image quality (1:1 pixel mapping) and simply launch a game from within Media Center, pick up my XBox game controller and play supported Vista video games using any PC resolution (without any overscan).
HT Slider
05-19-08, 07:23 PM
OK, this is getting even more bizarre...
When the ATI HDMI dongle is used, the UseBT601CSC=1 registry setting DOES change grey levels when watching DVDs and video within Media Center as well as Media Player.
Even stranger than that, to actually pass the BTB and WTW tests I have to crank up the overall CCC brightness not to just +15 or +16, but all the way to +31 and turn down the overall contrast to 74%.
To make things even more bizarre, with UseBT601CSC=1 set, no matter where I move the overall brightness and contrast I cannot pass Microsoft's grey level test patterns (wmv files within ehome directory). I can pass brightness (BTB) by cranking the brightness all the way to 61, but even with the contrast at 0 there is still clipping of bright objects (the shirt in the test).
The good news is using +31 for brightness and 74 for contrast makes my DVDs play perfectly as well as photographs, PC applications, desktop and video games all display perfectly.
When watching HD recordings in Media Center with +31/74, there still seems to be correct grey levels. I found some of the darkest material (blank frames, black material around credits, etc.) and brightest material and can't see any evidence of blacks showing up as grey or whites clipping.
When I use PowerDVD to watch Blue-ray, as long as I use the defaults for the AVIVO brightness and contrast and tweak the brightness and contrast within PowerDVD, it is still able to produce correct grey levels.
Ultimately, by using the ATI HDMI dongle, UseBT601CSC=1, +31/74% for overall brightness and contrast, default for AVIVO brightness and contrast, PowerDVD's advanced video settings for brightness/contrast to +20/-5 I get:
Photographs, PC applications and video games all with correct grey levels (0-255 visible grey levels)
DVD, SD & HD recorded TV with correct grey levels (16-235 visible)
PowerDVD Ultra with Blue-ray movies with correct grey levels (16-235 visible - note I can still fine tune this further if I had a Blueray calibration DVD)
The only thing that doesn't work (so far) is wmv playback (specifically Microsoft's brightness and contrast calibration clips) are completely out of whack with both far to little brightness and far to much contrast.
Since we don't watch much wmv content, so far this seems like the best calibration for our HTPC.
HT Slider
05-20-08, 12:16 AM
What a nightmare that was...
I installed the Nvidia 8500 GT and posted my experience here:
http://www.avsforum.com/avs-vb/showthread.php?p=13904784#post13904784
Unless someone can show me what I have done wrong, my next video card will DEFINITELY be an ATI again.
arfster
05-20-08, 03:57 AM
What a nightmare that was...
I installed the Nvidia 8500 GT and posted my experience here:
I used to have one of those, and can safely say NVidia's drivers are just as atrocious as ATI's. As you found, they've also recently started doing expansion by default - inevitable in a sense, given most video playback will be on on PC monitors, which by default are roughly calibrated to 0-255.
gbcrush
05-20-08, 08:41 AM
What a nightmare that was...
I installed the Nvidia 8500 GT and posted my experience here:
Unless someone can show me what I have done wrong, my next video card will DEFINITELY be an ATI again.
Wow, thanks. Actually reminds me a bit of when I was using my GF2 440mx to my 19" CRT in college. :D
I don't know whether you did something wrong or not, but it also made me realize I'm taking a lot for granted with my ATI card. Maybe I'll stick with it a little longer.
Thanks for posting.
HT Slider
05-20-08, 04:51 PM
I used to have one of those, and can safely say NVidia's drivers are just as atrocious as ATI's. As you found, they've also recently started doing expansion by default - inevitable in a sense, given most video playback will be on on PC monitors, which by default are roughly calibrated to 0-255.
I really don't mind expansion - in fact I would prefer it if I could see all of the appropriate grey levels between Media Center and PC applications (photographs, TV, Blue-ray, HD-DVD, games, etc.).
The key is I still want top image quality regardless of if the output is compressed to video levels or expanded to PC levels. With 10 bit per color processing there shouldn't be any quality loss so I say go for it - expand video and make everything view appropriately on any display.
JimmyFace
05-20-08, 09:21 PM
For Blu-ray calibration just use this:
http://www.avsforum.com/avs-vb/showthread.php?t=948496
It worked great for me. In PowerDVD I can't see the WTWs but I didn't try tweaking PowerDVD's controls, just the ones on my projector. I was leary of playing with the PowerDVD settings because I thought they may have more adverse affects then the CCC settings. But it seems that you've had luck so maybe I'll try tonight.
Great post, by the way.
HT Slider
05-20-08, 09:38 PM
For Blu-ray calibration just use this:
http://www.avsforum.com/avs-vb/showthread.php?t=948496
It worked great for me. In PowerDVD I can't see the WTWs but I didn't try tweaking PowerDVD's controls, just the ones on my projector. I was leary of playing with the PowerDVD settings because I thought they may have more adverse affects then the CCC settings. But it seems that you've had luck so maybe I'll try tonight.
Great post, by the way.
One thing to note is PowerDVD Ultra can only take over control of brightness and contrast if you keep the "let application decide" checkbox checked within CCC's AVIVO panel. If you alter the brightness or contrast in the AVIVO panel within CCC, the sliders do nothing within PowerDVD Ultra.
Thanks for the link. I'll see if I can create myself a couple of calibration DVDs that utilize the Blue-ray and HD-DVD playback engines within PowerDVD Ultra. (deleted my question after reading the link in more detail...)
arfster
05-21-08, 10:44 AM
Did a little test of screenshots to try various renderers, and the effect of expansion. Underlined two sentences that are essentially the results.
The frame used here is at minute 1:01:19 of the Bluray PotC2, which is AVC 1920*1080. This is displayed 1:1 on a 1920*1200 screen, and screenshot at the same in lossless PNG. In other words, no scaling at any point. As this is HD progressive, there is also no driver-based denoising/sharpening.
Setup: Sapphire 2600XT, Vista32, Cat8.4, latest ffdshow clsid build
Four screenshots:
A) ffdshow RGB32HQ > EVR
B) ffdshow RGB32HQ > EVR custom (3d/bilinear ps2)
C) ffdshow RGB32HQ > Haali renderer
D) ffdshow YUY2 > EVR, brightness 16/86 in CCC <-----ie standard drivers expansion, corrected back to video levels by CCC
With RGB32HQ done in ffdshow (A/B/C), you get no expansion at all, as long as you remember to set fullrange luma under RGB conversion (the default setting maps luma16 to rgb0, aka expands levels, see the sourceforge help page). Levels expansion appears to be always done at the RGB stage.
Some sample pixels:
X-Y
300-500 (dreadlocks) A/B/C 51/41/22, D 52/41/22 (same)
700-700: (overcoat) A/B/C 158/121/81, D 157/121/83 (D is -1 red, +2 blue)
700-300: (blue sky) A/B/C 148/180/205,D 149/180/204 (D is +1 red, -1 blue)
1600-400: (parrot body) A/B/C 240/180/43, D 239/180/47 (D is -1 red, +4 blue)
The various renderers produce literally 100% identical results (verified by image comparer util). As no scaling is happening, this is as it should be - they appear to be passing the decoder output through unmolested.
Comparing A/B/C with D, which has been expanded then contracted, I can't see any difference - good news. There are some minor differences when you look at actual pixels, mainly in the blue, but they're unnoticeable in practice even if you flick between the screenshots. They appear to be more the result of different RGB conversion than the expansion+contraction, because if you do a SD test (using usebt601csc to control expansion) then the differences are even more minimal, 1 rgb value at most.
Also did a test with the PDVD decoder, with and without acceleration, and both were 100% the same as D above - ie pdvd/ffdshow h264 decoders are producing the same output, whether hardware accelerated or not.
Zipped download here: http://rapidshare.com/files/116536001/screenshots.zip.html
More tests to come, perhaps some scaling/sharpening issues?
JimmyFace
05-21-08, 11:34 AM
HTSlider: I tested the PowerDVD controls and as soon as I touched the contrast the WTWs showed up, even at +1. In the end it was already calibrated properly because the optimal setting was contrast at zero, but I couldn't see the WTWs before by simply adjusting the projector's contrast. Thanks for the tip.
bobbyrr
05-21-08, 02:41 PM
8.5's just released.
Dave Baumann
05-21-08, 03:48 PM
Heres an overview of the the release note, some are relevent for here:
We have just released Catalyst 8.5 for Windows Vista, Windows XP and Linux platforms on http://www.amd.com
Highlights of the Catalyst 8.5 Windows release include:
Performance improvements
• Call of Juarez DX10 – Performance increases up to 12% on the ATI Radeon HD 3000 Series of products
• Lost Planet DX10 – Performance increases from 5 to 35% on the ATI Radeon HD 3000 Series of products.
• World in Conflict DX10 – Performance increases up to 25% on the ATI Radeon HD 3600 and ATI Radeon HD 3800 Series of products. Even greater gains are seen on the ATI Radeon HD 3870X2.
• Stalker DX9 – Performance increases by 20-50% when HDR is enabled in the game; across the ATI Radeon 3800 Series of products.
• Halo – Performance increases by 10-30% across all ATI Radeon graphics products.
New features
New display mode support
• This release of Catalyst™ introduces a number of display mode enhancements for the ATI Radeon HD 3000 Series, ATI Radeon HD 2000 Series, and ATI Radeon X1000 Series of products (configuration of these settings are adjusted in the Catalyst Control Center):
o SECAM TV-out support
o 1080p24 mode support
o 1080p Custom mode support for HDTVs (that support 60 Hz display modes) – users can configure their own custom 1080p modes to best fit their display.
Adaptive Anti-Aliasing – OpenGL support
• This release of Catalyst™ introduces Adaptive Anti-Aliasing support for OpenGL applications for the ATI Radeon HD 3000 Series, ATI Radeon HD 2000 Series, and ATI Radeon X1000 Series of products. Adaptive Anti-Aliasing substantially improves image quality by anti-aliasing transparent textures.
HDMI Audio for non-standard TV modes
• This release of Catalyst™ introduces support for HDMI Audio for all TV timing modes and non-standard (CEA 861b) TV modes. A dialog box will be shown to users when a non-standard TV mode is selected, asking if they would like to use HDMI Audio (in certain situations HDMI audio can get out of sync when running on a non-standard TV mode).
Windows XP Service Pack 3 support
• This release of Catalyst™ introduces WHQL certified support for Microsoft’s Windows XP Service Pack 3 update for all ATI Radeon Series products.
Resolved Issue highlights
• Clive Barker's Jericho: Corruption is no longer noticed when playing the game on a system containing an ATI Radeon™ HD 38x0 series of product and running the Windows Vista operating system.
• Crysis: Flickering shadow corruption no longer occurs in hallways of the ship when CrossFire™ is enabled.
• Gears of War: Launching the game under the 64 bit version of Windows Vista no longer results in momentary corruption of the desktop.
• Enemy Territory Quake Wars: Launch the game demo on a system containing an ATI Radeon™ X11x0 series of product and running the Windows Vista operating system no longer results in the operating system failing to respond.
• Hellgate: London: Flickering is no longer noticed when quitting and then relaunching the game when the in-game setting are set to their maximum levels.
• Stranglehold: The game no longer fails to load on systems running the Windows Vista operating system with default values enabled.
• Tabula Rasa: Setting the graphics option to enable Deferred Lighting and Dynamic Shadows no longer results in corruption being displayed.
• Team Fortress 2: Enabling CrossFire™ and playing the game on a system running Windows Vista no longer results in corruption being noticed.
• Tomb Raider Anniversary: Performance improvement is now noticed when playing the game with CrossFire™ enabled.
Highlights of the Linux Catalyst 8.5 release include:
New features
Catalyst Control Center Linux Edition – new 3D settings
• This release of the ATI Catalyst™ Linux driver introduces new Catalyst Control Center Linux Edition 3D settings for the ATI Radeon HD 3000 Series and ATI Radeon HD 2000 Series users. Linux users can now adjust the following 3D settings:
o Catalyst AI
o Mipmap detail level
o Custom AA filters
Resolved Issues
• Support for video playback on the second head in dual head mode is now available.
• Unresolved symbols in Xorg log under RHEL 4u6 32 and 64 bit no longer occur.
• The desktop background color is no longer displayed incorrectly when using xcommgr.
• Change VT no longer fails when kernel module is not loaded.
• Switching to the virtual terminal multiple times no longer results in X displaying corruption.
• Catalyst Control Center: The Preference page is now localized for all supported languages.
• The Linux kernel module is no longer installed to wrong location if the Linux kernel is updated.
• Random system hangs no longer occur when calls to the ATI driver are made from Ring3.
• X window no longer fails to respond when GoogleEarth is maximized.
• A segmentation fault no longer occurs when running SPECViewperf on systems containing an ATI FireGL V5100 series product and running SUSE 10.3 x86.
• Enabling Composite extension no longer displays 8x AA for graphic cards that do not support 8x AA under the Ubuntu 7.10 operating system.
• Display corruption is no longer noticed at logon when compiz is enabled.
• Launching the glxgears application no longer results in segmentation faults.
• GoogleEarth: Launching the application in horizontal mode on a system running X no longer results in corruption being noticed.
• Specviewperf 8.1 64 bit version: Segmentation faults no longer occurs on IGP systems.
• Maya 2008: Moving the render window no longer results in the operating system failing.
• Glxgears corruption is no longer noticed in big desktop mode.
For more information on Catalyst 8.5 (for Windows Vista, Windows XP, and Linux versions), including all of the resolved issues in this release, please see the Catalyst 8.5 release notes found in driver download section.
Anyone able to explain why 1080p via HDMI is considered new? Is it just because it is a pre-built mode now? Any chance 1080p over component is doable now?? That I would love (since my sharp takes 1080p via component).
Heres an overview of the the release note, some are relevent for here:
Dave, thanks for the heads up! Reading the announcement this could be a step into the right direction for HTPC users... :)
topcaser
05-22-08, 12:04 PM
Hi,
i did an update to XPs SP3, Realtek driver and updated CCC from 8.3 to 8.5 all at once.
I cant say that iam pleased with this sort of drivers. I get a lot of jaggies around the edges on SD content - with or without the Reg Hacks. Iam not sure where it comes from, but i will install 8.3 again.
BTW: Where can i download official 8.4 driver?
leeperry
05-22-08, 12:19 PM
do we finally get a pstrip-like in the drivers ?
I read that they offer custom resolutions creation ? :eek:
ATI_user
05-22-08, 01:09 PM
catalyst 8.5 UVD playback quality is poor than 8.2
Hi,
i did an update to XPs SP3, Realtek driver and updated CCC from 8.3 to 8.5 all at once.
I cant say that iam pleased with this sort of drivers. I get a lot of jaggies around the edges on SD content - with or without the Reg Hacks. Iam not sure where it comes from, but i will install 8.3 again.
BTW: Where can i download official 8.4 driver?
Be aware. I found that with 8.5 in contrast to all previous versions denoise / detail-enhance and all other postprocessing options are switch off by default.
grubi.
topcaser
05-22-08, 04:34 PM
@Grubi: And how is your SD content quality?
@Grubi: And how is your SD content quality?
Still evaluating. Hard to say at the moment. With default settings SD seems to be a bit pixelated, however as I had denoise activated previously I cannot be compared.
arfster
05-23-08, 08:09 AM
Could problems people are having be down to 8.4's broken uninstall? I had to manually clean it out because it left bits all over (even more than usual).
Anyway, I'll do some screenshot comparisons between different drivers (thinking pre-sliders in particular). However, does anyone know how to make rollover comparisons? Or where can host them?
Luar Azul
05-23-08, 08:53 AM
By the way, your image quality probs with coreavc are probably down to it using different levels (config, select pc or video/tv levels). It shouldn't look any different other than that.
Today I had to use CoreAVC for a 1080p file that just didn't work with MPC-HC decoder. Even configuring the input-output levels, the picture quality was disappointing (black levels seemed fine, but the picture was very pixelated). Later on I tried with other files with which I compared coreavc with mpc hardware decoder, the result was clear: With coreavc there is a lot of pixelation in faces and same color surfaces (like walls). When I use MPC decoder I get less pixelization, less bothering, closer to film grain. I had "standard deblocking" selected on coreavc and I'm using version 1.7.
With MPC decoder and 8.4 ati drivers I can see almost everything with excellent quality. Amazing how a freeware program works way better than the paid decoder from CoreAVC or even mammoth Cyberlink. :) Thanks to the MPC-HC team!!!!!! :D
leeperry
05-23-08, 09:19 AM
arfster : the drivers "enhancements" don't show up on screenshots, as they are processed after the mixer.
Strayshot
05-23-08, 09:40 AM
FWIW, I installed 8.5 last night using my normal routine of ATI uninstall util=>safe mode=>driver cleaner=>reboot=>install normal. I had no problems installing. CCC came up just fine. All of my deinterlacing options were there. I set to VA, checked the pulldown, and tested on HQV SD disk. I see no functional difference from 8.4. The flag scene looked fine - no stepped edges, no pixelization. The film detail (racetrack with bleachers) deinterlaced fine. Noise reduction seems to be working well too, according to the HQV test.
Anyway, I am one who has no problem with these drivers.
XP SP3
2600XT PCIe
P35 chipset
arfster
05-23-08, 10:11 AM
arfster : the drivers "enhancements" don't show up on screenshots, as they are processed after the mixer.
Are you meaning the driver-based sharpening? (ie CCC sliders). That's definitely captured by screenshots.
If Haali vs driver/EVR, I tried two ways to look for uncaptured differences: play simultaneously and alt-tab between them on the same screen, and simulanteously on two screens (my "workstation" has two of the same 1920*1200 screens). Other than minor colour differences (bt709 differences presumably), they were identical - definitely no postprocessing going on there.
Of course, that's just my setup. As with anything else related to HTPC, what you actually end up with is down to your individual hardware/software setup+ the phase of the moon + a few other random factors.
ps upscaling might (and probably is) show more considerably differences, but the above tests were intended to bypass all scaling by displaying at 1:1.
one_2go
05-23-08, 10:42 AM
Did a little test of screenshots to try various renderers, and the effect of expansion. Underlined two sentences that are essentially the results.
Setup: Sapphire 2600XT, Vista32, Cat8.4, latest ffdshow clsid build
Four screenshots:
A) ffdshow RGB32HQ > EVR
B) ffdshow RGB32HQ > EVR custom (3d/bilinear ps2)
C) ffdshow RGB32HQ > Haali renderer
D) ffdshow YUY2 > EVR, brightness 16/86 in CCC <-----ie standard drivers expansion, corrected back to video levels by CCC
With RGB32HQ done in ffdshow (A/B/C), you get no expansion at all, as long as you remember to set fullrange luma under RGB conversion (the default setting maps luma16 to rgb0, aka expands levels, see the sourceforge help page). Levels expansion appears to be always done at the RGB stage.
Thanks for your diligent efforts in getting the best out of the ATI cards. IN what way do these findings change the playback chain? Is the ffdshow RGB32HQ > EVR a suggestion to use ffdshow as decoder or in post processing OR was this just for the test putposes?
Thanks for any help. Just got a MSI HD3650 and want to try the 8.4 drivers with this new card with built in HDMI connector.
arfster
05-23-08, 10:53 AM
Is the ffdshow RGB32HQ > EVR a suggestion to use ffdshow as decoder or in post processing OR was this just for the test putposes?
You can do either for this test - I did it with ffdshow as a decoder cos it's simpler.
Anyway, the purpose of the test was to find the ideal setup for displaying TV levels. At present, that means recontracting (brightness16/contrast86) after the default expansion ...... messy! I'd done this a while back with SD, but HD has some differences (bt709 for one). The ffdshow approach I used does the YV12>RGB32 conversion in software, rather than the standard setup of the renderer doing it in hardware, and thus avoids levels expansion completely.
Anyway, the conclusion is basically that for TV levels, expansion+contraction is not 100% lossless, but it's only minutely off. Blindtesting, I don't think anyone could see the difference, except perhaps on a specially picked and very artficial test pattern (and it would be minimal even then).
The ffdshow approach I used does the YV12>RGB32 conversion in software, rather than the standard setup of the renderer doing it in hardware, and thus avoids levels expansion completely.I don't think this is true. On my set up - 3850 Cat.5.4 on XP with dongle - doing RGB conversion fixes the BT.601/BT.709 problem and I get video levels but I still have clipped BTB and WTW. So it seems I am still getting expansion and contraction. Needless to say I am not doing any level adjustment in ffdshow.
HT Slider
05-23-08, 12:53 PM
Anyway, the conclusion is basically that for TV levels, expansion+contraction is not 100% lossless, but it's only minutely off. Blindtesting, I don't think anyone could see the difference, except perhaps on a specially picked and very artficial test pattern (and it would be minimal even then).
Keep in mind FFDShow probably only uses 8 bit colors.
If you let ATI do the video processing in hardware, it uses 10 bits per color plus there is no handoff between directshow filters. Compressed video is sent to the card and it takes it all the way to displaying the final result on the screen. In reality it doesn't even need to convert between YCbCr and RGB (depending on how the software is written).
Do you find the image quality is better with FFDShow? Unless it is doing a better job of processing the video, why wouldn't you simply take advantage of the hardware processing on your 2600?
The problem is that unless you use ffdshow to do the conversion you get the wrong colorspace conversion from the driver if you have upscaled the video in ffdshow.
The driver assumes that all video 720p and above is BT.709. SD (e.g. DVD) is BT.601. But if SD has been pre-upscaled the driver does not know to use this.
The colorspace errors are far worse than any errors due to 8-bit rounding. ffdshow HQ RGB32 conversion is very good (if slow!).
HT Slider
05-23-08, 01:10 PM
The problem is that unless you use ffdshow to do the conversion you get the wrong colorspace conversion from the driver if you have upscaled the video in ffdshow.
The driver assumes that all video 720p and above is BT.709. SD (e.g. DVD) is BT.601. But if SD has been pre-upscaled the driver does not know to use this.
The colorspace errors are far worse than any errors due to 8-bit rounding. ffdshow HQ RGB32 conversion is very good (if slow!).
I must be missing something.
If I don't use FFDShow at all, don't even have it installed, use Media Center for my HD recordings and PowerDVD Ultra for my HD-DVD/Blueray, use an ATI HDMI dongle on the card output plus separately fine tune both the ATI brightness/contrast and PowerDVD brightness/contrast to dial grey levels right in - am I not getting an image at least comparable to using FFDShow?
How do you use FFDShow for protected Blueray and HD-DVD anyway? Can PowerDVD use it if AnyDVD HD is installed?
I totally agree colorspace errors are far worse than rounding errors. I am trying to understand what the advantage to FFDShow is over DXVA AVIVO.
arfster
05-23-08, 01:12 PM
I don't think this is true. On my set up - 3850 Cat.5.4 on XP with dongle - doing RGB conversion fixes the BT.601/BT.709 problem and I get video levels but I still have clipped BTB and WTW. So it seems I am still getting expansion and contraction. Needless to say I am not doing any level adjustment in ffdshow.
Not using the dongle, but there's no clipping going on for me, see full 0-255 here:
http://img156.imageshack.us/img156/1939/rampus1.th.png (http://img156.imageshack.us/my.php?image=rampus1.png)
Have you set ffdshow rgb input to fullrange? The default setting expands, and perhaps then the dongle is causing recontraction?
Not using the dongle, but there's no clipping going on for me, see full 0-255 here:
http://img156.imageshack.us/img156/1939/rampus1.th.png (http://img156.imageshack.us/my.php?image=rampus1.png)
Have you set ffdshow rgb input to fullrange? The default setting expands, and perhaps then the dongle is causing recontraction?Just trying to make sure I understand you here:
At the start you are saying that you are not using the dongle, right?
What is this about rgb input? ffdshow is getting YV12 from Nvidia MPEG-2 filter and is then outputing RGB. Only filters are blur and resize.
arfster
05-23-08, 01:20 PM
Do you find the image quality is better with FFDShow? Unless it is doing a better job of processing the video, why wouldn't you simply take advantage of the hardware processing on your 2600?
The idea was to avoid expansion+contraction to get TV levels, thus skipping an unnecessary stage and keeping btb/wtw. That worked, but it turned out the extra stage was visually lossless.
I don't use ffdshow upscaling so bt601/709 issues there aren't a factor, although I do have to for the occasional 720p mkv that is encoded without bars (thus is approx 550ish pixels high), and isn't flagged as bt709 - the drivers then assume it's SD, and use bt601.
I must be missing something.
If I don't use FFDShow at all, don't even have it installed, use Media Center for my HD recordings and PowerDVD Ultra for my HD-DVD/Blueray, use an ATI HDMI dongle on the card output plus separately fine tune both the ATI brightness/contrast and PowerDVD brightness/contrast to dial grey levels right in - am I not getting an image at least comparable to using FFDShow?
How do you use FFDShow for protected Blueray and HD-DVD anyway? Can PowerDVD use it if AnyDVD HD is installed?
I totally agree colorspace errors are far worse than rounding errors. I am trying to understand what the advantage to FFDShow is over DXVA AVIVO.None of this applies to Blu-ray or even high def (except the case of 2.35:1/2.4:1 720p video encoded without the black bars as arfster explained) . I am talking about using ffdshow to upscale SD. ffdshow can still produce a subjectively better (sharper, without artifacts) upscaled image than DXVA AVIVO, at least in the opinion of many observers(!). But upscaling confuses the driver and if you pass it YUV causes the wrong colorspace conversion. So purely in the case of pre-upscaled SD you need to do RGB conversion in ffdshow.
arfster
05-23-08, 01:22 PM
Just trying to make sure I understand you here:
At the start you are saying that you are not using the dongle, right?
What is this about rgb input?
Sorry, edited it too fast, meant RGB conversion (the tab, under output). Under contrast section defaults to standard, which expands.
Not using the dongle, no.
Sorry, edited it too fast, meant RGB conversion (the tab, under output). Under contrast section defaults to standard, which expands.
Not using the dongle, no.??? "Standard" in my copy of ffdshow says 16-235 luma, 16-240 chroma (latest stable beta4).
There is another option below for "full-range" 0-255.
Mine is set to "standard".
Resultant image is clearly clipped as a result of expansion (presumably by the ATI driver), although the levels are right due to the dongle.
arfster
05-23-08, 01:33 PM
??? "Standard" in my copy of ffdshow says 16-235 luma, 16-240 chroma.
There is another option below for "full-range" 0-255.
Mine is set to "standard".
That's expanding. It's confusingly named, but those are actually input luma ranges. It always maps output to RGB0-255, so if you select standard/16-235 you get expansion.
Try flicking between them when playing something, standard is darker than fullrange.
That's expanding. It's confusingly named, but those are actually input luma ranges. It always maps output to RGB0-255, so if you select standard/16-235 you get expansion.
Try flicking between them when playing something, standard is darker than fullrange.
Oh! OK. That's clear!:confused::)
I'll have a go. Anyway, it is fine, because I want video levels and the dongle is doing it's contraction thing, so I need expansion first, but it is good to know what is doing it. Thanks!
I'll have a go.You are right. It was ffdshow doing expansion to "0-255" due the the "standard" "luma 16-235" setting :rolleyes::)
arfster
05-23-08, 01:49 PM
Yeah, stupidly named isn't it :-) That had me tearing my hair out for a bit.....
Anyway, nice to know the dongle contracts in that situation, whereas brightness/contrast in CCC doesn't affect video when you feed the renderer RGB.
As an aside, I wonder if dongle-contraction avoids chroma problems (ie does it have the equivalent of the modify-only-luminance tickbox in ffdshow's levels?).
Edit: I guess it does when you use ffdshow for rgb conversion, but how about normal decoders outputting yuy2/yv12? Urrrgh, I'm hungover, this is making my head hurt.
Yeah, stupidly named isn't it :-) That had me tearing my hair out for a bit.....
Anyway, nice to know the dongle contracts in that situation, whereas brightness/contrast in CCC doesn't affect video when you feed the renderer RGB.
As an aside, I wonder if dongle-contraction avoids chroma problems (ie does it have the equivalent of the modify-only-luminance tickbox in ffdshow's levels?).
Edit: I guess it does when you use ffdshow for rgb conversion, but how about normal decoders outputting yuy2/yv12? Urrrgh, I'm hungover, this is making my head hurt.I believe it does the right thing. Hard to be sure unless you capture images on the far side of the HDMI cable. But, subjectively, when I was doing these tests:
http://www.theatertek.com/Forums/showpost.php?p=67600&postcount=67
http://www.theatertek.com/Forums/showpost.php?p=67601&postcount=68
there was no difference in RGB vs DXVA YV12 (live video of course, not the screenshots which would all be the same). The chroma problems of BT.601/Bt.709 stood out a mile.
I obviously cannot tell if ALL have a subtle chroma problem without removing the dongle and re-calibrating, my eye is not that good, but I doubt it.
arfster
05-23-08, 02:14 PM
Makes sense I think.
Thinking backwards, the only way you can have chroma problems here is if it does the contraction on the YUY2/YV12 before RGB conversion (cos if it's done afterwards, it's operating on RGB and there's no separate luma).
Easy to test: feed the renderer yuy2, check for clipping. If there is any, dongle-contraction is happening after the expansion in RGB conversion (cos contraction before expansion would maintain btb/wtw), and thus there's no potential problem.
STEVEKEYS
05-23-08, 08:01 PM
Setup:
- HD 2600 xt connect to Onkyo via HDMI
- Onkyo 805 receiver
- Playing blu ray on Dell PC
- Using Cyberlink Powerdvd Ultra to play blu ray disk
- Windows XP SP 3
The video looks great and I get sound, but I'm not getting the Dolby TRUHD that's on this Blu ray disk passing to the receiver via HDMI....I'm just getting 2 channel stereo(the 805 is still in Dolby PLII mode when the blu ray disk is playing). In my "Sounds and audio device properties" in XP, under the "audio" tab, I have "ATI HD Audio rear output" set. In Powerdvd ultra,under the "audio" tab, I have "use s/pdif" set under speaker environment. Under "output mode", I have "only primary audio" set...the only 2 other options here are AC3 mixing and Dts mixing. Tryed both and still don't get the truehd passed to my receiver. Tryed a few other blu rays I own and still no surround modes are passed to my Onkyo(yes I selected them).
I'm positive this blu ray has TrueHD for I when I play this disk on my PS3, it decodes the TrueHD perfectly. The other blu rays I've tried have 5.1 uncompressed, 5.1 dolby digital, etc., and can't get these disks to pass the surround audio via HDMI either using the 2600 HD XT.
Any ideas would be appreciated!
originalsnuffy
05-23-08, 11:37 PM
If I recall, we have previously established that the 2400 and 2600 can only pass 2 channel pcm via hdmi, along with bit reduced formats like regular dolby digital or dts.
This is a 2400 and 2600 design limit, not a hdmi limit.
HT Slider
05-24-08, 12:56 PM
I thought I'd post one final update to my grey level struggle, final grey level success and my trial with the Nvidia 8500GT.
If any of you read through the 8500GT trial post (http://www.avsforum.com/avs-vb/showthread.php?p=13904784), you'd see that I found absolutely no way at all to get consistent grey levels using the 8500GT. To get the Nvidia to produce a decent image I had to literally tweak the Nvidia control panel before switching applications absolutely all the time. This included switching between Media Center, PowerDVD, Media Player and Desktop/Video games. Nvidia doesn't even provide hotkey capability for Vista so I literally had to exit my application, pull out my mouse and keyboard, start the control panel and manually make all the changes. The Nvidia card also did a terrible job of upconverting 480i content into 1080i, while the ATI card does a great job of this.
Last night I reinstalled the 2600XT along with Catalyst 8.5 and have to say even with all of the hassles of setting it up, I am SO much happier with the 2600XT vs the 8500GT. I have now, finally got absolutely ALL grey levels working correctly (for the applications I use) and automatically. I can simply start PowerDVD from within Media Center, start a video game from within Media Center, exit to the desktop, view photographs, etc. Everything works and I can do everything simply using the remote control (and, of course, plugging in the XBox 360 game controllers into the front of the HTPC for gaming)!
On top of that interlaced video output (1080i) is much better with the ATI. The Nvidia did a great job at 720p, but for some reason 1080i flickered much worse than the ATI. With the Nvidia, the desktop wasn't even legible at 1080i, but with the ATI is is (although still clearly interlaced).
Here are the settings and configuration that work with my particular HDTV, the Toshiba 51H83, and my HD 2600XT video card (note that I suspect my HDTV does something strange with the video card and requires different settings than typical):
Note that I also took advantage of JimmyFace's Blu-ray/HD-DVD calibration link (http://www.avsforum.com/avs-vb/showthread.php?t=948496) and both the Blu-ray (AVCHD) and HD-DVD calibration DVDs worked flawlessly with PowerDVD Ultra.
HDTV hooked up using the ATI HDMI dongle (this is critical - actually each and every one of these points is critical).
ATI Catalyst 8.5 installed (over 8.4, suddenly wmv playback became compliant when using these settings here).
UseBT601CSC=1 added to the DXVA portion of the Catalyst video driver.
Overall brightness and contrast set to +31 and 74% respectively.
AVIVO brightness and contrast set to default with the check-box selected to let applications take control.
PowerDVD Ultra's advanced video settings for brightness and contrast set to +19 and -5 respectively (calibrated using the calibration disks mentioned above).
Even though there is absolutely no way this level of tweaking should be required to get compliant video output, I am still very pleased to FINALLY have correct grey levels for everything I use.
In addition to that, in order to get the best image quality I also (adjusted while using calibration disks where possible):
Created "custom resolutions" in the HDTV control panel within CCC that perfectly fill the screen for both 720p and 1080i (ATI also uses these overscan compensated resolution for all PC resolutions, but still outputs spec compliant 720p and 1080i when using 1280x720 or 1920x1080).
Calibrated my HDTV after calibrating the video card. Note that while calibrating the video card I cranked the brightness way up and contrast way down to ensure the HDTV wasn't limiting any of the grey levels being visible (this way I will have the greatest number of grey levels available, or no unnecessary grey level compression). I also discovered that as a final test my DVD calibration disk played back using the hardware DVD player confirms that my HDTV is also now calibrated for it too (also note that I adjusted the brightness and contrast using the service menu so the default setting of 50 is now correct in the user menu for brightness and contrast). I also have flesh-tone correction turned off and auto contrast off.
Turned down the sharpness on the HDTV to 30 (default is 50 - I didn't look for sharpness in the service menu).
Turned off ATI's flesh-tone and color vibrance using the control panel.
Set the AVIVO deinterlacing to automatic with pulldown detection enabled (default).
Set the AVIVO noise filter at 40 (note since decreasing the sharpness on the HDTV I have found new settings I consider best for the noise filter and edge enhancement; I used to use 50/15 with default sharpness and now prefer 40/20).
Set the AVIVO edge sharpening at 20.
I now have the following grey levels visible as well as excellent image quality:
Photographs using Media Center: 0-255 visible
Photographs using Vista's photo viewer: 0-255 visible
Video games: 0-255 visible
Desktop: 0-255 visible
DVD in Media Center: 16-235 visible
DVD in Media Player: 16-235 visible
Recorded SD & HD TV in Media Center: 16-235 visible
Recorded SD & HD TV in Media Player: 16-235 visible
Blu-ray in PowerDVD Ultra: 16-235 visible
HD-DVD in PowerDVD Ultra: 16-235 visible
WMV in Media Center: 16-235 visible
WMV in Media Player: 16-235 visible
Absolutely every single piece of content I want to display and piece of software I want to use is now producing both correct grey levels and excellent image quality. On top of that, the HDTV's overscan is perfectly compensated for automatically when I use any standard PC resolution or my custom resolutions; yet still spec compliant 720p and 1080i (including overscan) is output when 1280x720 or 1920x1080 is used. Nvidia doesn't support anywhere near this level of overscan compensation either.
Also, due to the fact that ATI uses 10 bit color accuracy for all of their processing, I am convinced there is effectively no degradation due to potential expansion and contraction of grey levels.
I am sure many will appreciate your detailed post on your experiences.
But I am baffled by the need to tweak brightness/contrast at all either in CCC or PDVD. It sounds like your TV's calibration is not to video levels.
All the other steps I concur with and produce perfect video levels and chroma in all apps with my TV & a 3850 (still on Cat 8.4). I am currently using TheaterTek, PDVD7, PDVD8, MPC-HC + photos. Someone else with a 2600 has also got perfect results with the same.
I would suggest others start with the tweaks, minus brightness adjustments, and get into those if needed.
I am sure many will appreciate your detailed post on your experiences.
But I am baffled by the need to tweak brightness/contrast at all either in CCC or PDVD. It sounds like your TV's calibration is not to video levels.
All the other steps I concur with and produce perfect video levels and chroma in all apps with my TV & a 3850 (still on Cat 8.4). I am currently using TheaterTek, PDVD7, PDVD8, MPC-HC + photos. Someone else with a 2600 has also got perfect results with the same.
I would suggest others start with the tweaks, minus brightness adjustments, and get into those if needed.
HT needs to add in support for the MS decoder in VMC. He will no doubt explain but my guess is that, as there are no controls for this decoder and CCC is set to "Application control", he needs to calibrate for VMC playback then use PowerDVD's settings to adjust for HD content.
there's one thing I don't understand. I thought that PDVD settings didn't take effect with CCC set to "application controlled". Is this not the case?
Wo0zy
HT Slider
05-24-08, 01:53 PM
I am sure many will appreciate your detailed post on your experiences.
But I am baffled by the need to tweak brightness/contrast at all either in CCC or PDVD. It sounds like your TV's calibration is not to video levels.
All the other steps I concur with and produce perfect video levels and chroma in all apps with my TV & a 3850 (still on Cat 8.4). I am currently using TheaterTek, PDVD7, PDVD8, MPC-HC + photos. Someone else with a 2600 has also got perfect results with the same.
I would suggest others start with the tweaks, minus brightness adjustments, and get into those if needed.
I don't get it either. My HDTV passed all of the grey level tests using a hardware DVD player both prior and after doing all of this calibration with the ATI 2600.
If I don't touch anything at all (install CCC and use default settings), DVD, WMV and SD Recorded TV playback is 100% correct when using either Media Center or Media Player. Unfortunately that is where things end.
Photographs, desktop, video games, PowerDVD, HD Recorded TV, etc. is totally screwed up. With photographs, desktop and video I can only visibly see grey levels from 16-235. With HD Recorded TV, I can see something like 30-220. With PowerDVD I can only see something like 30-220 also.
Once I add the UseBT601CSC=1, it changes all SD content, including DVD, WMV and SD Recorded TV. At that point it takes the overall brightness cranked up to +31 and overall contrast turned down to 74% before DVD, WMV, and SD Recorded TV is back to spec. Luckily this also fixes Photographs, desktop, and video games (0-255 is now visible).
PowerDVD Ultra is strange in that it somehow ignores the overall CCC brightness and contrast settings. I can crank up/down the AVIVO brightness and contrast to +15/85% to get PowerDVD to spec, but this breaks video playback grey levels in Media Center and Media Player.
Luckily, instead of adjusting the AVIVO settings in CCC, I found I can use PowerDVD's settings instead. Using +19 and -5 gets me bang on with perfect 16-235 grey levels (16 being black, 235 being white).
HT Slider
05-24-08, 02:05 PM
Actually, let me make one correction to my above explanation.
If I add the UseBT601CSC=1 and instead use AVIVO's brightness/contrast at +15/85 I can fix DVD, SD Recorded TV, WMV as well as PowerDVD Ultra's grey levels. Unfortunately this leaves my desktop, photographs and video games with only 16-235 visible.
Using the complete method outlined in my "big post" above, I can get absolutely everything to provide correct grey levels (in other words 16-235 visible for all video and 0-255 visible for desktop, photographs and video games).
UseBT601CSC=1 was already used in your first post. Do you mean UseBT601CSC=0?
Wo0zy
HT Slider
05-24-08, 02:18 PM
All the other steps I concur with and produce perfect video levels and chroma in all apps with my TV & a 3850 (still on Cat 8.4). I am currently using TheaterTek, PDVD7, PDVD8, MPC-HC + photos. Someone else with a 2600 has also got perfect results with the same.
When you say + photos, what do you mean?
Are you saying that pictures at the desktop are shown with the full 0-255 grey levels visible? What about photos within Media Center and video games (should be the same - at least it is on mine)?
Are you certain you have correct grey levels everywhere while using totally default settings in CCC and without using any registry hacks?
I have managed to assemble quite an assortment of calibration source content throughout all of this. Maybe I'll try to post links to everything so everyone can confirm they are actually seeing 16-235 for all video content and 0-255 for all PC type applications, video games and photographs.
Based on what others have said it does seem possible that somehow my HDTV produces different grey levels between different applications when compared to other systems. I'm not sure if my HDTV somehow uses different grey levels between RGB and YCbCr sources, if other enthusiasts are not calibrating all of the sources, or if there is a difference between our calibration sources.
At this point I am quite confident that I do finally have the right grey levels visible everywhere - for the first time ever.
HT Slider
05-24-08, 02:29 PM
UseBT601CSC=1 was already used in your first post. Do you mean UseBT601CSC=0?
Wo0zy
No, I do mean UseBT601CSC=1.
I'll start again...
Default settings using the HDMI dongle:
DVD, SD Recorded TV, WMV - all correct with 16-235 visible.
Everything else wrong (including HD TV, photographs, PowerDVD Ultra and video games).
Adding UseBT601CSC=1:
Everything wrong, with video far too dark with too much contrast.
If I now adjust the AVIVO brightness/contrast to +15/85%:
Most, if not all video is now correctly calibrated for 16-235 (including PowerDVD)
Unfortunately photographs, desktop and video games also only have 16-235 visible at this point
If I instead adjust the AVIVO brightness/contrast back to default and check "application controlled" + crank up the overall brightness to +31 and overall contrast to 74% in CCC:
Photographs, desktop, and video games all have 0-255 visible (good)
DVD, SD Recorded TV, HD Recorded TV, and WMV all have 16-235 visible (even better)
Unfortunately PowerDVD Ultra is wrong (bad)
If I now crank up the brightness to +19 and contrast down to -5 in PowerDVD Ultra (advanced video settings), I now get absolutely everything working.
Does this make sense?
How does this compare to what others are seeing with your HDTVs?
I understood your first post perfectally. It was the use of the word "instead" in the follow-up that seemed odd. I guess you meant "also".
Sorry.
Nice work BTW!
Wo0zy
HT Slider
05-24-08, 03:12 PM
I understood your first post perfectally. It was the use of the word "instead" in the follow-up that seemed odd. I guess you meant "also".
Sorry.
Nice work BTW!
Wo0zy
Is my experience in-line with what you are seeing? (assuming your display is an HDTV - if it a PC monitor, this probably changes everything)
I'm wondering if the values I converged on are specific to my Toshiba 51H83 or if they work for all displays?
HT Slider
05-24-08, 03:22 PM
HT needs to add in support for the MS decoder in VMC. He will no doubt explain but my guess is that, as there are no controls for this decoder and CCC is set to "Application control", he needs to calibrate for VMC playback then use PowerDVD's settings to adjust for HD content.
there's one thing I don't understand. I thought that PDVD settings didn't take effect with CCC set to "application controlled". Is this not the case?
Wo0zy
BTW, your guess was 100% correct.
PDVD settings do work with CCC AVIVO brightness/contrast set to "application controlled". They don't work when the AVIVO brightness is specifically set in CCC.
One thing I discovered is if an Nvidia card is used (Nvidia drivers), PDVD can't control the brightness or contrast no matter what. The sliders "slide", but they don't do anything. This is also how PDVD behaves when the AVIVO "application controlled' check box is not selected.
I am using PDVD 3730a OEM.
BTW, your guess was 100% correct.
PDVD settings do work with CCC AVIVO brightness/contrast set to "application controlled". They don't work when the AVIVO brightness is specifically set in CCC.
One thing I