[Vorbis] oggenc -q switch vs -b switch

[gt]

On Mon, Jun 20, 2011 at 11:28:35PM +0530, gt wrote:
> I have filed a bug report here:
> 
> http://sourceforge.net/tracker/?func=detail&aid=3323303&group_id=86862&atid=581181

Well the bug report was rejected by the mediainfo author, and here's the
reason he gave:

----------------------------------------------------------------------

zenitram(Jerome Martinez) said:

Vorbis header has this:
0000001C Identification (30 bytes)
0000001C Signature: 1 (0x01)
0000001D Signature: vorbis
00000023 Version: 0 (0x0)
00000027 Channels: 2 (0x02)
00000028 SamplingRate: 44100 (0xAC44)
0000002C BitRate_Maximum: 0 (0x0)
00000030 BitRate_Nominal: 224000 (0x36B00)
00000034 BitRate_Minimum: 0 (0x0)
00000038 BlockSize_0: 11 (0xB)
00000038 BlockSize_1: 8 (0x8)
00000039 Framing: 1 (0x01)

Currently, MediaInfo handles such info as a "goal" of 224 Kbps, i.e. a
kind of ABR (Average Bit rate), not a VBR (with 2 pass analysis, 1 st pass
to detect needs, 2nd pass for coding high differences)

I looked a bit at the bitstream, each packet is between 4000 and 4500 byte
long, each granule position difference is around a value of 8000, variation
is very very tiny... So I see no reason to set it as VBR (or I should also
set all CBR MP3s as VBR! MP3 has a bit reservoir in each CBR frame --> this
is not pure CBR too...), I put VBR when I detect more differences (extrem
example: bitrate at 16 Kbps during 10 seconds then 448 Kbps during 10
seconds, average of 224 Kbps too and this is really a VBR stream)

----------------------------------------------------------------------