Log in

View Full Version : DVD Experiment to Test Buffer Underflow Theory


Peter Cheat
3rd October 2004, 11:47
I would like some volunteers who have had playback troubles with a particular DVD thay have authored. This includes trying to fix jerky playback and out of sync audio. It will required remuxing the file and (maybe) burning it again. In particular, I was to test whether crappy playback is caused by the maximum bitrate being exceeded, or a buffer underflow. Any volunteers?

Sir Didymus
3rd October 2004, 16:07
Hi Peter!
:)
I am sorry your preface is excluding myself... since I never had any playback trouble (at least in recent Rebuilder releases...).

On the other hand I am very curious and proactive for all the tests aimed at pointing out any (very few and sporadic IMHO) remaining bugs in the beautiful program of jdobbs. So maybe you accept anyway my posting into your thread...

If you could explain better what you have in mind it would be appreciated...

For the moment my contribution is just related to the following:

1. I think it is possible to exclude that the bitrate in itself is the cause of crappy playback. This has been already demonstraded, in other threads. Once the movie has been encoded, the resulting bitrate is one of the properties of the encoded MPEG stream. Some other kind people (if I remind right) have already reauthored, with Scenarist or other tools, some m2v streams produced after the encoding step of DVD-RB obtaining the elimination of the dropouts they experienced using DVD-RB.

2. As a personal feeling, STD buffer overflows or underflows are more convincing; but this is just a feeling, so (IMHO) much more testing should be performed for (a) giving proofs that these anomalous conditions hold and (b) for understanding if these eventual conditions are really related to the observed problems (and that they are not just formal and ininfluent flaws in the streams produced by DVD-RB) and (c) most important, for giving (if possible) some precise indication on what exactely is the cause of these anomalies...

In the meanwhile we'd better wait for the announced 063, and see if somebody is still having problems with unpreprocessed titles...

There is nothing wronger than to worry about some very marginal (or already solved) problem...

Just as a curiosity, are you actually experiencing "crappy playback" ?

Cheers,
SD

robot1
3rd October 2004, 23:17
Originally posted by Peter Cheat
I would like some volunteers who have had playback troubles with a particular DVD thay have authored. This includes trying to fix jerky playback and out of sync audio. It will required remuxing the file and (maybe) burning it again. In particular, I was to test whether crappy playback is caused by the maximum bitrate being exceeded, or a buffer underflow. Any volunteers? Probably you just need to use a picky and serious Authoring program, like Scenarist.
If I'm not wrong, when rebuilding Scenarist complains if the stream could cause a VBV buffer underflow.
So it's not necessary to use rebuild, then remux and author... if you trust Scenarist (or another Tool).

jdobbs
3rd October 2004, 23:42
My suggestion: If anyone experiences "crappy playback" they should use something else to copy DVDs. The vast majority of people don't, and that's the reason DVD-RB has become so popular.

@Sir Didymus

Download version 0.63. I spent the last few days doing extensive tests with "runt" VOBUs and I've found them to be the cause of several possible problems. I've fixed it in this version. the vulen program you provided really came in handy.

@Robot1

This version also supports your EclPro program with Canopus Procoder version 1.5 and above... you may want to test it.

As for bitrate overruns -- I also did much, much testing with that and frankly it is a red herring for all but a very few, very rare instances -- but just to protect against it, this version adjusts the max bitrate downward as selected audio bitrate goes higher.

robot1
3rd October 2004, 23:53
Originally posted by jdobbs
@Robot1

This version also supports your EclPro program with Canopus Procoder version 1.5 and above... you may want to test it.
Testing now, it seems ok.
I'll post a link to EclPRO soon.

jdobbs
4th October 2004, 00:04
Download version 0.63. I spent the last few days doing extensive tests with "runt" VOBUs and I've found them to be the cause of several possible problems. I've fixed it in this version. the vulen program you provided really came in handy. Let me make this clearer -- I found it to also DEFINITELY be the cause of several errors.

Thanks much, Sir Didymus, for your continued support. This was a really good catch.

dragongodz
4th October 2004, 00:06
i better explain what this is about.

Peter and i disagree on what is the acceptable max bitrate for dvd video. i say its 9800Kb per second for both reading and decoding. he says no thats just the read limit from the dvd drive while a stream exceeding that is fine on input to the VBV buffer, so basically the max bitrate is not important only the VBV.

i mentioned that earlier DVD-RB had some cases where max bitrate spiked above the 9800Kb/sec and could cause non-smooth playback. as jdobbs said himself he adjusted the max bitrate setting that DVD-RB uses down so it should not be a problem now. that however does not change what we are talking about. :)

best to go have a read here
http://forum.doom9.org/showthread.php?s=&postid=552403

jdobbs
4th October 2004, 00:51
:cool:

I read the topic... interesting arguments. My own take is that 9,800Kbs is stressed over-and-over as the maximum video bitrate. The CCE manual has a section in it in which it mentions the fact that the buffer can handle input bursting to 11Mbs and still be MPEG compliant -- while it is debatable whether it remains DVD compliant. CCE will limit bursts to 9.8Mbs when the "DVD compliant" flag is set -- so I think their take is the same as yours. Since the standard isn't clear here, it's less risky that way.

Peter Cheat
4th October 2004, 03:00
I was hoping someone whould have a problem that could not really be pinpointed, as the stream seemed ok when muxing, bitrate viewer didn't report anything out of the ordinary, but the file played back incorrectly nevertheless. I believe some software used for muxing does strange things, and I wanted to test this out. I can't fix buffer underflows (could possibly be done by manually changing data in the stream (like with ReJig)?). For VBR, the maximum bitrate can be exceeded, and shouldn't cause problems. I think that some muxing software may not like it if you go over the bitrate indicated as the max, causing playback problems.

Sir Didymus
4th October 2004, 11:07
Originally posted by jdobbs
Download version 0.63. I spent the last few days doing extensive tests with "runt" VOBUs and I've found them to be the cause of several possible problems. I've fixed it in this version. the vulen program you provided really came in handy.


Just dowloaded.

Thanks a lot for all the hard work you are putting into DVD-RB. You should be aimed by some very strong motivations into this project!!!

Thanks also for your very kind words. I am fully rewarded by them, by using DVD-RB :D and by the understanding of being helful, in some way.

I will continue testing (and using!!!) your wanderful application...

All the best,
SD