|
|
#1 | |
|
Registered User
Join Date: Mar 2007
Posts: 45
|
Hello,
With both the 1.0.9755 and the 100.14.03 drivers I get a strange issue whenever I play 1080i content at 1080i. This is on a CRT RPTV connected via DVI (I also have this problem via component). The problem appears to be a syncing issue whenever 1080i content is played at 1080i. I am guessing it is a syncing issue. I have no way to verify. It's on a timer apparently. The problem will persist for 15 seconds and then go away for 15 seconds and then will come back for 15 seconds etc... 720p and 480i/p content will look just fine. The only way I know how to describe it is blurriness. Parts of the picture (typically around action) appear to jump or jerk. I can observe the problem with both mythtv and mplayer. When connected via DVI I use a custom modeline since I have the problem discussed in this thread. With component connections I don't use the custom modeline. I have attached a bug report. It includes the xorg.conf I use when I am using the DVI cable. |
|
|
|
|
|
|
#2 | |
|
Registered User
Join Date: May 2007
Posts: 2
|
Hi,
Interlaced modes play content the way you described when frame speed is not exactly synced, mplayer for example determines playback speed by evaluating responses from the audio driver. As a workaround you may use the -speed option in mplayer, set it to a value a bit larger or smaller than 1.00000, depending wether video swaps too fast or too slow. AFAIK, mplayer has no option to determine its playback speed by evaluating the Vertical Retrace Interrupt (not that I really know if nvidia drivers are passing them to applications). |
|
|
|
|
|
|
#3 | |
|
Registered User
Join Date: Mar 2007
Posts: 45
|
Quote:
I'll try the -speed option. I believe MythTV has an option to sync using the audio (although it's not really recommended). Thank you for the reply. |
|
|
|
|
|
|
#4 | |
|
Registered User
Join Date: May 2007
Posts: 2
|
Hi,
It is not your TV that is unusal. The TV should adapt to any speed the video card supplies (within the tolerated range). The problem is that there is no internal communication in mplayer regarding synchronization with the video card. For smooth playback to work, either - the video card would have to swap frames when requested to do so by mplayer (kind of internal genlocking), which is not possible with nvidia hardware (s.o. correct me if I'm wrong), or - mplayer would have to listen to Vertical Retrace Interrupts from the nvidia hardware, and adjust its playback speed, which it doesn't do. Instead, it listens to the audio hardware, to provide gapless audio playback. I don't know who is the culprit for the inaccuracy on my system, but either of the quartzes of the video card, audio card, or mainboard has a tolerance of more then 0.2% (which is 1 frame in about 20 seconds, similar to your experience, if my math is correct). The problem gets worse when displaying live feeds, such as DVB streams, because both audio and video hardware would have to be sync'ed to the input speed. If you don't force mplayer to play SLIGHTLY slower than input speed, you will sooner or later run into buffer underruns. Complicated topic. The best solution is to use a de-interlacer with full progressive frame interpolation, so out-of-sync fields don't show up as jagged movements. But the -vf=yadif filter is too slow on my machine for 1080i content, I use it for 576i (PAL) content though. |
|
|
|
|
|
|
#5 | |
|
Registered User
Join Date: Aug 2003
Posts: 1,026
|
Quote:
We get interlaced content, de-interlace it in the video player to place it in a progressive video frame store, which is again sent to the TV in interlaced way. The TV usually converts it to progressive again (for 100Hz scanning or LCD/Plasma display). What we get is a very sub-optimal picture quality. There are many things that can go wrong, including getting the field order wrong (combining the wrong fields into a frame). And then there are the hiccups. Even with progressive video all the way, there is the synchronization issue when there is a free-running framestore in between (as is the case in a PC videocard). Sad. This must be the reason why video playback via a computer is so inferior in quality to using standard DVB and DVD equipment directly connected to a TV. |
|
|
|
|
|
|
#6 | |
|
Registered User
Join Date: Dec 2005
Posts: 37
|
Playing 1080i video with 1080i output USED to work fine circa 8183 driver version. The 'sync' bug/problem was introduced afterwards.
It isn't just mplayer. Mythtv, xine and other video players have this problem. One way of describing the problem: Pause the video output, you'll noticed the driver will display interlaced fields very jittery. Very noticeable during horizontal panning scenes. |
|
|
|
|
|
|
#7 | |
|
Registered User
Join Date: Aug 2003
Posts: 1,026
|
Quote:
You don't see this when the player de-interlaces the video, but when the interface between PC and TV is again interlaced this probably will degrade the picture as it is de-interlaced twice. |
|
|
|
|
|
|
#8 | |
|
Registered User
Join Date: Mar 2007
Posts: 45
|
Quote:
|
|
|
|
|
|
|
#9 |
|
Registered User
Join Date: Mar 2007
Posts: 45
|
A question: some people have this problem (1080i content not syncing properly when using 1080i out), but it would seem that a lot of people don't. What is the difference between these two sets of users? How is that some people don't seem to have a problem playing 1080i content while, say, I do?
One possible reason is that the people without the problem simply don't see it because they are de-interlacing (it was suggested earlier in this thread that de-interlacing 'resolved' the syncing problem). I'd like to point out that when I use de-interlacing of 1080i material @ 1080i I don't have the syncing problem, but I have numerous other problems depending on the deinterlacing technique that I use. Regardless I never can get a decent picture for 1080i content @ 1080i. |
|
|
|
|
|
#10 | |
|
Registered User
Join Date: Aug 2003
Posts: 1,026
|
Quote:
|
|
|
|
|
|
|
#11 |
|
Registered User
Join Date: Dec 2005
Posts: 37
|
I understand what pe1chl is referring to. What I am trying to describe is so much more 'blatant' of something wrong with the interlaced fields. It is hard to describe, but it is the source of the problem. The timing or sync problem is causing fields to either 'stay' longer or display 'faster' than they should.
As to you saying others have successful working 1080i content on 1080i output, I have yet to find one person to do so successfully under certain conditions such as 1080i content via 1080i DVI. Some combination of using the older drivers with VGA output with a VGA->Component converter works fine, but 1080i over DVI has never worked 100 percent. The older drivers had problems displaying any video via XV/XVMC over DVI/Component output at 1080i resolution but did not have the same problems via VGA. The newer drivers made it possible to play video with XV/XVMC over all the outputs but messed up with the 1080i output, therefore requiring deinterlacing (even thru the VGA port). Deinterlacing is certainly an option to obtain a 'useable' display, but not the optimum. There just isn't a lot of people using 1080i output. Pretty much only CRT based HD sets use it. LCD/Plasma sets are usually 720p or 1080p, or people are using a computer monitor. A quick and easy test would be for Nvidia to get their hands on a 1080i CRT set and use the DVI/HDMI port to play 1080i content. It would be easily noticeable, assuming they can get the 1080i resolution to be accepted without mucking with modelines. If it works 100 percent fine for them, then just post the benchmark system information such as motherboard/chipsets, video card, linux distro and xorg.conf config. |
|
|
|
|
|
#12 | |
|
Registered User
Join Date: Aug 2005
Posts: 47
|
Quote:
After adding the Option "UseEvents" "True" to my xorg.conf and enabling OpenGL vertical sync in MythTV I got my 1080i artifacts reduced very significantly, but still not quite good enough for me to disable deinterlacing. With 1080i content, anything that pans fast can get really bad. It's really a shame...this is pretty much the only flaw I have in a really great (and expensive) system. Tom |
|
|
|
|
![]() |
| Thread Tools | |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| NVIDIA Responds to Reports of Kepler V-Sync Stuttering Issue | Rieper | NVIDIA GeForce 600 Series | 13 | 03-03-13 10:56 PM |
| Sync it up: Hands on with the preview of Windows 8's cloud sync service | News | Latest Tech And Game Headlines | 0 | 06-02-12 08:30 PM |