Experimental TsReader development (2 Viewers)

[Owlsroost]

Microsoft video codecs, ffdshow audio codecs, mpaudiorenderer. Maybe it's because of the last one. I hope Tourettes will pick up developing it. Can you?
I'll try to figure out which one gives the best results 33 or 34.

I've also found that MP AudioRenderer doesn't seem to like large jumps in the timestamps (e.g. from bad signal corruptions) - that's why I asked which audio renderer you were using

It shouldn't mid about corrupted / dropped data. All it does is it schedules the 1st sample after the discontinuity point to be rendered in time when the rtStart tells. Of course there could be some bug, but at least I haven't noticed such during the 1,5 years the renderer has been in use.

Was the timestamps modified somehow? Maybe a backwards jump?

wouter1971 - when it goes out of sync, is the audio early or late compared to the video ?

(I'm not volunteering to try and fix it, but it might be useful info for anyone who does....)

Tony

 

[wouter1971]

It shouldn't mid about corrupted / dropped data. All it does is it schedules the 1st sample after the discontinuity point to be rendered in time when the rtStart tells. Of course there could be some bug, but at least I haven't noticed such during the 1,5 years the renderer has been in use.

Was the timestamps modified somehow? Maybe a backwards jump?

Tourettes, i believe have read that you don't have a 5.1 setup. So personaly you don't use the AC3 encoding mode i guess?
I think that's where this issue is.

 

[tourettes]

It shouldn't mid about corrupted / dropped data. All it does is it schedules the 1st sample after the discontinuity point to be rendered in time when the rtStart tells. Of course there could be some bug, but at least I haven't noticed such during the 1,5 years the renderer has been in use.

Was the timestamps modified somehow? Maybe a backwards jump?

Tourettes, i believe have read that you don't have a 5.1 setup. So personaly you don't use the AC3 encoding mode i guess?
I think that's where this issue is.

Yep, I dont use AC3 re-encoding. But there is no timestamps anymore at that point, so I doubt the issue is specific to the AC3 re-encoding. It is quite easy to test, just use ffdshow as audio decoder (set it to downmix to stereo) and then disable the AC3 re-encoding.

 

[wouter1971]

wouter1971 - when it goes out of sync, is the audio early or late compared to the video ?

(I'm not volunteering to try and fix it, but it might be useful info for anyone who does....)

Tony

Tony, i appriciate all the help you give us. Let me check this, i had corruptions this evening as well and had to skip back -15s twice to get it in sync again on those moments. I believe the sound comes earlier than the video and it is about a second delay.

 

[wouter1971]

Hi Tony,

Tested with latest tsreader 35 and LAV as decoder a stall in the last minute. All logs attached.
Still testing on the audio/video sync with mpaudiorenderer and the playback of the recording of last nights show in which i got the corruption during live tv with audio/video sync issue.

Greetings,

Wouter

 

[wouter1971]

Tony,

Is tested with the tsreader 35 andthe playback of lastnights recording in which i got the corruption and audio/video sync problem.

Tswriter (last nights recording):

08-12-2011 19:57:11.921 Recorder: RECORD start of audio detected
08-12-2011 20:22:20.747 Recorder: Pid 202 Continuity error... f ( prev a ) - bad signal?
08-12-2011 20:22:20.748 Recorder: Pid 202 Continuity error... f ( prev a ) - bad signal?
08-12-2011 20:22:20.748 Recorder: Pid 22 Continuity error... 2 ( prev 8 ) - bad signal?
08-12-2011 20:22:20.748 Recorder: Pid 22 Continuity error... 2 ( prev 8 ) - bad signal?
08-12-2011 20:22:20.748 Recorder: Pid 54 Continuity error... d ( prev e ) - bad signal?
08-12-2011 20:22:20.748 Recorder: Pid 54 Continuity error... d ( prev e ) - bad signal?

Tsreader (today's playback):

09-12-2011 15:19:34.287 [664]Video Continuity error... f ( prev a )
09-12-2011 15:19:34.287 [664]Audio Continuity error... d ( prev e )
09-12-2011 15:19:34.726 [12ac]aud:set discontinuity
09-12-2011 15:19:35.701 [1330]vid:set discontinuity

Evr.log (today's playback):

09-12-2011 15:19:36.063 [135c]Setting media type on mixer
09-12-2011 15:19:36.064 [135c]SetupAudioRenderer: cycleDiff: -0.0000069378
09-12-2011 15:19:36.064 [135c]SetupAudioRenderer: adjust bias to : 1.0000069378
09-12-2011 15:19:36.064 [135c]SetupAudioRenderer: Delayed Audio by : 0.0000000000
09-12-2011 15:19:36.260 [7a0]Dropping sample from the past (6.55 ms, last call to NotifyWorker: 0.20 ms, Queue: 2, Dropped: 25)
09-12-2011 15:19:36.268 [135c]ProcessMessage MFVP_MESSAGE_INVALIDATEMEDIATYPE
09-12-2011 15:19:36.268 [7a0]ProcessOutput: change of type


What i'm experiencing is that last night the LiveTV was out of sync after the corruption (missing tsreader from that moment).
But today's playback, tried it about 6 times and a few times they same part to check but audio never went out of sync after the corruption.

Looks like there is a difference between LiveTV playback and recorded playback?

Full logs attached when i was repeating playback of that sample about three times.

 

[cruse]

AW: Experimental TsReader development

Sorry for the delay, here are the logs for v35 and stock tsreader.ax. Both done with LAV 0.42. AudioRenderer DirectSound (stock).

v35
about 90% CPU, at the beginning video corruption, later more stuttering and audio desynch

stock
Sorry dont shoot the messenger but no corruption at all. Just the usual "aftertune stutter" about 30-45 seconds in and a small stutter in the end, which might be due to my OSD harassment which i turned on and off at this time to make MP sweat. Around 85% CPU.

 

[sukceso]

AW: Experimental TsReader development

After some testing, I have not seen the "hang" problem with the new version. Thank you very much for investigating and improving TsReader. I don't see any corruption or stuttering, works very well in my environment.

 

[technick]

FYI LAV 0.42 have also problems with videos (suddenly video stopping during the movie), personally I return to 0.39 It will be difficult to test Tsreader if decoder (and sometime dshowhelper) change also.

 

[Owlsroost]

FYI LAV 0.42 have also problems with videos (suddenly video stopping during the movie), personally I return to 0.39 It will be difficult to test Tsreader if decoder (and sometime dshowhelper) change also.

Thanks for the info

Tony