Version 1.2.1 introduced new rice coding techniques that are used by the reference encoder for 24 bit files. An older version of the decoder will have trouble with frames that use this encoding... Maybe that's where the strange noises come from...
Pyt. On 6 févr. 2011, at 06:01, Brian Willoughby <bri...@sounds.wa.com> wrote: > Thanks for bringing up this aspect, Nicholas. I seem to recall that > specific hardware has a problem with certain compression levels, but > I cannot recall whether that was limited to just encoding, or > decoding as well. It could very well be true that I am conflating my > vague memory of encoder limitations with decoder limitations. > > It does seem to be that the oppo BDP-95 is exhibiting problems with > particular flac files. Since my original message, my friend has > installed the latest version of flac and recompressed the exact files > that were giving him a problem before - now with -0 or --fast he > doesn't see a playback problem at all. So, even though your > statements make total sense to me, the evidence seems to indicate > something about the compressed data that's causing a problem. The > original audio is not the issue, but how it is compressed. > > Here's a thought: Since the encoder looks for polynomials, could it > be possible that certain decoders cannot handle certain polynomials > in real time? > > Ah, another possibility is that the oppo BDP-95 implements an older > version of the decoder, and it's merely new flac files that give it a > headache. My friend happened to have an old version of flac > installed on his computer, 1.1.4, and that reported stream errors > with his files until he upgraded to 1.2.1 - if the oppo has anything > older than 1.2.1 then I suppose that might explain the decoding > problems. > > Brian > > > On Feb 5, 2011, at 16:33, Nicholas Wilson wrote: >> Correct me if wrong, but I was under the impression that the >> processing required for playback was totally independent on the >> level of compression. The encoder looks for polynomials that fit, >> and it takes much more processing to find polynomials with a very >> good fit and small residuals. On the other hand, the decoder just >> has to multiply out the stored prediction, which is independent of >> the compression level. >> >> Nicholas > _______________________________________________ > Flac mailing list > Flac@xiph.org > http://lists.xiph.org/mailman/listinfo/flac _______________________________________________ Flac mailing list Flac@xiph.org http://lists.xiph.org/mailman/listinfo/flac