Audio buffermodel...

[Enots_]

In the DVD docu version 1.11 on page VI5-32 Table 5.3-1 Input Buffer for the player reference model it states that the input buffer size of audio is 4kB. However for example on page VI5-24 (AC3 PES-header) the P-STD_buffer_scale is 1 and the P-STD_buffer_size 58 giving according to the P-STD docu (ISO/IEC13818-1 /ITU-T rec H.222.0 1995 E page 40) a 58k Buffer for such a substream. Now this leaves me a bit confused - how much input buffer must a DVD player now have for audio?

[jj59]

Hi,
dunno what DVD docu you are referring to but aren't you talking about MPEG-audio vs audio contained in a Private_Stream_1 like AC3, DTS and LPCM. Because a Private_Stream_1 can also carry subpictures (SPU)the buffer size is a lot bigger for this type of streams.
Some more info is here
http://forum.doom9.org/showthread.php?s=&threadid=81115

regards
jj

[Enots_]

I was refering to "DVD Specifications for Read-Only Disc / Part 3. Video Specifications Version 1.11"

I think you have a point that for audio packs (MPEG) the buffer is 4Kb for the reference player which matches what documentation says about the PES-extension about this kind of pack. Unfortunately there is a footnote on the audio field in table 5.3-1 where I have my 4 kB. "Remove unit = one audio frame plus private headers, if any (*3)" ... "*3: private headers is defined as follows
1) In case of Linear PCM, private header consists of following items.....
2) In case of Dolby AC-3...
3) In case of MPEG-1 audio or MPEG-2 Audio...
4) In case of DTS (option),.....
5) In case of SDDS..."

If audio was refering only to MPEG there would never be any private headers and the sentence would be entirely meaningless. If it refers to AC3 (and/or the others audio types in privatestreams) we have at best an inconsistency about what the true input buffersize of a DVD-player must be for these types of streams.

I have a tendency to believe that the PES-headers are right and that table 5.3-1 must be wrong. 4kB input buffer would make muxing of LPCM awfully difficult when there is many streams and near 10.08 sys-bitrate.

So the true question is a) Am I missing a point somewhere? or b)Which of the two interpretations are correct? Anybody got any empirical evidence as to which is true?