Win Intel HTPCs/NUCs & Kodi-native 3D MVC Playback
(2016-04-11, 10:08)noggin Wrote: Is this with bitstreamed audio for the three different formats, or with audio decoded to multichannel PCM?
One 23.976p frame has a duration of 41.7ms, so we're looking at between 0.5 and 4.5 frames of A/V delay.
All these numbers are for bitstreamed audio. Comparable AV sync error was present even with decoded multichannel PCM. That kind of surprised me. Yes, we are looking at 4-5 frame delay in some cases. You can actually quite easily notice this with the test pattern.

(2016-04-11, 14:13)hdmkv Wrote: @wesk05, I'm wondering the same thing, whether there's A/V sync delay w/LPCM as indicated by Sync-One2, and if so, to what degree(s)? Also, not sure I saw which video types the tests were with (3D ISO's)? Since it appears there is A/V sync delay no matter what w/bitstreaming, it's a question of what's perceptible (and becomes annoying) to each of us. Instruments help w/unbiased data, but w/o them, most of us will need to rely on good 'ol eyes 'n ears Smile.

I am finding it odd there's such variance between different builds though, so hopefully further development can address this for most Intel h/w.
The numbers in the table are all for 2D test pattern and bitstreamed audio. I got somewhat similar numbers with decoded PCM audio also. I haven't yet done a complete test with decoded audio. I have done some measurements with 3D test pattern also. The numbers weren't much different from those of 2D. Yes, the perceptible error will be different for each individual. I start to get annoyed only when it exceeds ~100ms. I will PM you a link for the 3D ISO version of the test pattern. I would appreciate if you checked it on your systems. With @afedchin's builds, I get a green line at the bottom. I don't get this on the Pi2. So, I am not sure whether it is a problem with my file or not.

(2016-04-11, 03:40)wizziwig Wrote: Have you tested any other builds besides this experimental MVC version using this method? I wonder how much of that a/v sync error is just inherent to Kodi in general. Would be interesting to see if non-Windows like OpenElec does any better, including on devices like Rpi.
I have actually tested it on nVIDIA SHIELD and Pi2. I think most of it is inherent to Kodi and also system (AVR, display) specific. SHIELD was interesting in that over time (~1 min) the error got corrected by itself. Pi2 also had ~140ms error (bitstreamed and decoded audio).



I do want to mention here that the numbers that I have posted are not absolute. They are specific to my system. I did the tests only to check the variation between different Kodi builds.
Reply
 
Thread Rating:
  • 10 Vote(s) - 4.3 Average


Messages In This Thread
RE: Intel-specific HTPC/NUC Selection w/reliable MVC 3D ISO - by wesk05 - 2016-04-11, 19:04
What about xvYCC / x.v.Color - by ma9ick - 2016-04-15, 14:01
June 30 download link broken? - by CooperCGN - 2016-08-02, 08:18