> X-Authentication-Warning: geek.rcc.se: majordom set sender to
>[EMAIL PROTECTED] using -f
> X-Sender: [EMAIL PROTECTED]
> Date: Fri, 09 Jun 2000 17:01:20
> From: Shawn Riley <[EMAIL PROTECTED]>
> Content-Type: text/plain; charset="us-ascii"
> Sender: [EMAIL PROTECTED]
> Precedence: bulk
> Reply-To: [EMAIL PROTECTED]
> X-UIDL: 2eG!!OpQ"!mZ8"!?@n!!
>
> I suspected that "free format" meant that the encoder wasn't
> restricted to the standard bitrate options, so if I'm right about
> that & if an MPEG Layer3 stream was freeformat & had no bit
> reservoir, would it be natively VBR? Do you recommend the "--nores"
> option for VBR mode? Also, could MP3 be made lossless by using a
> masking threshold of -96dB for all frequency bands & allowing larger
> framesizes than 320kBit/sec? (Ignoring that it could end up taking
> more data than the PCM equivalent to do so...)
>
> Shawn
> --
Hi Shawn:
1.
free format with no bit reservoir would be natively VBR and does seems
much more logical then what is used in the MPEG spec :-) If the
convention of no ancillary data (extra data that is not decoded) used
in free format bitstreams was adopted, this would be easy to do. But
I think the problem comes from the fact that MPEG wanted to allow
ancillary data: In a free format VBR bitstream, you would have to read
through the (possible) ancillary data after decoding each frame,
looking for the next sync word. if the ancillary data accidently
contained this syncword, it would confuse the decoder.
2.
--nores should only be used for testing and the special case of
producing streams which can be cut and pasted on frame boundaries.
It will result in some loss of quality.
3.
MP3 cant be made lossless: The MDCT/filterbanks are not
perfect, although the errors are quite small, and there will
always be some quantization error since the MDCT coefficients
are converted to integers between 0 and 8192 for encoding.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
