Scenarist & bitrate too high (yet again)

[redguy]

hi everybody
i searched the forum but i have not found an answer fitting to my problem.
when i build the dvd i'm authoring this message pops up



Error Video or Audio Buffer underflow: (Dts: 1102, SCR 1170)
Error dvd_mux : DoMux Multiplexing Error
Error Terminated Multiplexing (NTSC_8dvd9-03-despertar NTSC-t_t.vob).
Error Total bitrate is too HIGH. Please reduce the stream bitrate or the number of stream.
Error Multiplex is failed.
Error Multiplexing failed, Track "03-despertar NTSC-t"

Error MuxFromDB:MuxVTS failed
Error Multiplex failed
Error DVD Video files could not be created


Strange enough since i took care not to surpass the maximum bitrate allowed, so I verified the track in scenarist itself

Info === Verification 03-despertar NTSC-t ===
Info -- Total bitrate --
Info Main video bitrate = 7500000, data = "03-despertar NTSC"
Info Audio stream ID = 1, bitrate = 1536000, type = PCM, data = "pcm2.0 03-despertar"
Info Audio stream ID = 2, bitrate = 512000, type = AC-3, data = "ac35.1 03-despertar NTSC"
Info Sum of audio bitrate = 2048000.
Info Total bitrate = 9548000 [ audio + main video ]
Info === Finish.[Verification] ===


as you can see i have not exceeded the maximum bitrate. to be sure of it i rechecked the video stream in bitrate viewer and it yielded a maximum bitrate of 6283 (I originally encoded VBR 1500-6000-7500).

:confused:

This track is the 3rd chapter of the main show. It is encoded just like the other two but those build OK and this does not. Soooooooo, puzzled, i decided to test just this track alone in a new project and it built OK.


what'am i doing wrong???

[mpucoder]

What went wrong is somewhere the video bitrate peaked too high for the mux.
Let's clear up a few things:
1) What Scenarist reports for video bitrate is taken from the video streams sequence header, it does not reflect reality (as you can see, it matches your maximum bitrate parameter exactly).
2) Bitrate viewers (any of them) do not analyse video bitrates from a multiplexer's point of view, ie they only show the maximum bitrate per picture. But multiplexing does not work on a per picture basis, but on a sliding window of approximately (depending on the muxer) 7 frames working with a 232K buffer. At no time can there be more than 232K in the buffer, so when bitrates start going up and the buffer gets full, the multiplexer has to shut off loading the buffer. This causes the data to get delayed, and increases the bitrate from then on until lower bitrate pictures allow the system to catch up. Bottom line - if the 6 or so frames before a 6000K picture fill the buffer and cause delaying the data for the 6000K picture, it could take 12000k or more to transfer the picture's data before the data is needed for decoding. This being impossible causes a mux failure.

The fix is always the same - lower the bitrate, at least for the section that cannot mux.

[redguy]

I think I understand. So you can have set correctly your maximum bitrate (from the encoder point of view) but still you can ruin your mux if there are many consecutively pictures with high bitrate.

This can explain the fact that I can mux it alone in the track editor, or in a brand new project and not with the other tracks, so the problem must lie perhaps in the frames' bitrate of transition from track 2 to 3.
Am I right?

Is there must be some kind of protection against this problem at the encoder stage?

[mpucoder]

If you can mux track 3 seperately with no problem then you might want to consider set track 3's first cell to non-seamless. This has the effect of isolating the tracks, making them essentially independant muxes, but at the price of a pause in playback on most players.

[redguy]

i see but i can't do that. i have reencoded the 5.1 audio down to 448kbps and it worked. dammit. i should manually lower the bitrate at the beginning of track 3 and the ending of 2.

thanks mpucoder!! your wisdom amazes me

[bigotti5]

Info Audio stream ID = 1, bitrate = 1536000, type = PCM, data = "pcm2.0 03-despertar"
Info Audio stream ID = 2, bitrate = 512000, type = AC-3, data = "ac35.1 03-despertar NTSC"
Info Sum of audio bitrate = 2048000.

DVD compliant AC-3 bitrate is 96–448 kbps

[mpucoder]

I read that somewhere, too. But in reality Scenarist and Muxman both support rates as high as 640K, and players have no problems with these rates. I think the restriction was on version 1.0 of the spec. And the reason was that rates higher than 448 have frames larger than the payload of one pes packet, and therefore produce packs with no frame header. But there now is an accepted way to multiplex audio frames > 2025 bytes.

[DaRat]

I got this exact same problem while i was working on a mixed angle project, it was caused by a bitrate peak in the previous segment (not the segment scenarist was screaming bloody hell about). I guess in the light of MPUCoder's post it makes sense now. :D