Hello,

My question will sound stupid (and probably is) but Im a complete newcomer in the audio field.

In computer music compression systems are MP3, OGG, FLAC an so on

In DVD / Blu Ray compressions systems are Dolby, DTS...

How are they related ? what are the differences between these two family ?

Thanks

MP2/MPC are related subband codecs.
MP3 is a hybrid subband/transform codec.
OGG, WMA and AAC are all very similar transform codecs that evolved more or less at the same time in response to mp3.
FLAC is a lossless format like zip or rar.




On DVD they're both older, somewhat inefficient lossy codecs. AC3 is transform, I don't know about DTS. I don't know what the new bluray flavors are like.



Formats used on disks tend to have much higher bitrate and support for very large numbers of channels. They also tend to use simpler, less efficient compression and less CPU time so that they're cheaper to implement on DSP cores.

Thanks a lot mike !!!

Luc

Not really. MP3 uses two levels of sub-band coding, one using the PQMF from MP2, and the other the misnamed "modified discrete cosine transform" which is, in fact, a critically sampled, 1:1 and onto (orthonormal, tight frame, what-have-you) SUBBAND analysis system with a whole lot of desirable properties, but not when coupled in tree structure with another filterbank. There are basic theorems that show that tree-structured filterbanks have some interesting, but often not so useful, properties regarding time/frequency tradeoff.

No, no, and no.

OGG was developed to "work around" patents. I won't discuss that issue. It uses the MDCT and is a high-band-count subband coder. More I can't say.

AAC is a descendent of AT&T Bell-Labs PAC with a few tidbits from MP3 thrown in, and Temporal Noise Shaping added. It uses a 1024-point or 128 point (switched) MDCT sub-band analysis bank. It handles up to 48 channels per bitstream. It was intended as a new algorithm in place of MP2/MP3 for new uses in the MPEG community.

WMA is a high-band-count Sub-Band coder for which the details are not entirely well known.
WMA-Pro is an advanced, better version of WMA that handles multichannel and low bit rate scenarios.

They did not evolve "at the same time", nor did they evolve "in response to mp3".

There are, I must note, very few actual transform coders reported. The original "PXFM" coder was a transform coder, as were several others (CNET?), all of which switched to sub-band coding via MDCT in their more advanced versions.

AC2 is a medium-band-count MDCT coder.
AC3 is a medium-high band count MDCT coder that handles more channels.
Older DTS were low-band-count coders, I'm not sure what the present situation is.

While it may seem pedantic to argue about "transform vs. suband", in fact, sub-band coders are critically sampled, and do not cause data growth. In audio coding, transform coders can not be critically sampled. Some Sub-band filterbanks (MDCT) are "exact reconstruction" in the mathematical sense, and some (the PQMF in Layers 1,2, 3) are not exact reconstruction in that sense, but all are critically sampled. By comparison, transform coders must have overlap, or alternatively they must have block-rate artifact problems. Th eMDCT is both 1:1 and power complimentary. Some SBC filterbanks are exact reconstruction, but are not power complimentary (i.e. not tight frames), for instance biorthogonal filterbanks. Some are simply not exact reconstruction (PQMF, QMF).

On the other hand, a transform coder obeys Parsival's theorem on a block by block basis, at least in the fiterbank, and a sub-band coder can not. This is good for psychoacoustic modelling, but not so great for coding thanks to the oversampling in the frequency domain.

Sorry if this is too much information, but I figure that too much can be ignored, and too little may not give you the handle you want to examine things.

I hear the reference encoders for libvorbis beta 4, RC 2, and 1.0 are all very different, so YMMV.