Talk:MPEG-1 Audio Layer II

From Wikipedia, the free encyclopedia

This is an interesting article, but there are few things I either don't understand or perhaps don't agree with.

The statement that sub-band coders work in the time domain, while MP3 is a transform coder and works in the frequency domain could be inaccurate. Surely both MP2 and MP3 are sub-band coders in that they divide the frequency space to be encoded into a number of bands. There are several ways of doing this. One way is to use a set of filters, and then it should certainly be possible to operate in the time domain as suggested. However the input would also have to be divided up into time slots, with a decision made as to which filter+time slot combination to encode. Essentially each time slot+filter combination has to be associated with a number of bits for encoding, which 0 bits implying that the combination is not encoded at all.

This processing is clearly not linear.

MP3 works in pretty much this way, and my understanding was that the reason that MP3 gave greater compression than MP2 is because MP3 uses temporal masking as well as frequency masking. With temporal masking low level signals both before and after a masking signal can be removed - apparently.

For implementing filters it is nowadays computationally effective to use a DFT algorithm, rather than implement a set of filters. I suspect that both MP2 and MP3 do this. Speech data compression used to be done by filter banks until eventually it became obvious that using DFT was feasible, and more effective. It is possible that some audio compression tools were being developed before the use of digital filtering became widespread and DFT could be done effectively in real time. For example early developments in DCC (Digital Cassette Tape) and MiniDisc could perhaps have been based on filters rather than DFT processing. Some very early systems may even have used analogue filters. This is perhaps of historical interest.

There is a statement that MP2 uses 32 channels, while MP3 uses 576. This may be true - I don't know. It does seem to me though that since MP3 is capable of encoding a wide range of sampling frequencies and quantisation levels, that it may not actually use a fixed set of channels. In other words, does MP3 always use 576 channels, or is the number of channels somehow dependent on the input signal. There might, for example, not be much point in using 576 channels for voice input sampled at 8kHz.

The observation that at high bit rates MP2 is somehow "more accurate" than MP3 is interesting, though it's unfortunate that many broadcasters in the UK use lower bit rates, with only Radio 3 consistenly broadcasting at 192kbps. This is just another example of a not so well known observation, which goes like this:

Codec A is better than Codec B at low frequency data rates. What can we say at higher data rates? Answer: NOT MUCH - indeed PERHAPS NOTHING AT ALL

Some software purveyors have tried to convince us that a particular codec may be better by demonstrating that it does better at a given data rate. It does not necessarily follow that at different data rates the same properties will hold. For example, Microsoft's WMA codec is clearly better (I really believe this ...) than MP3 at rates of 48kbps and lower. This says absolutely nothing about their relative quality at rates of 96kbps and above, and compatibility factors would tend to favour MP3. This is because different algorithmic features may kick in at different compression rates. aacPlus and mp3Pro both use SBR (Spectral Band Replication) at low data rates, which enhances them relative to AAC and MP3, but at higher bit rates they probably just converge to plain AAC and MP3.

David Martland 07:51, 27 December 2005 (UTC)

---

I just had a look through this article and have a few comments:

Both MP2 and MP3 split the audio into different frequency sub-bands. After this MP3 uses a MDCT to resolve each sub-band into frequency components. Depending on the block length of the MDCT, this can either result in 192 or 576 frequency components. See http://wiki.hydrogenaudio.org/index.php?title=MP3

So, in response to above, both MP2 and MP3 do use filters, but only MP3 uses the MDCT. See also: http://www.cs.columbia.edu/~coms6181/slides/6R/mpegaud.pdf

The comments about Ogg Vorbis do not have citations and are too vague. There is no mention of what artifacts the glockenspiels caused or why Vorbis would avoid them. Vorbis still employs the MDCT, making it similar to many other codecs. Vorbis has also had its issues with artifacts, namely pre-echo, although these were really only a problem with the older encoders.

The explanation of why MP3 provides better compression than MP2 is just plain wrong. Again, please see the above pdf for a description of the design improvements of MP3 over MP2.

[edit] ""incorrectly" called Musicam"

Musicam is the name used for MP2 in the specifications for DAB and Astra Digital Radio as well as in the BBCs DAB documents - it may not be a common name these days for MPEG 1 Layer 2 audio compression, but it seems to be common enough in older documents, leading me to think its not "incorrectly" called Musicam. --Kiand 02:18, 3 February 2006 (UTC)

Support. I deleted the corresponding passage. --Abdull 17:55, 7 March 2006 (UTC)

[edit] Links to MP3

Sorry if this has previously been mentioned, but why is history on MP2 stored in the MP3 article. (Just look at the top of this article and there is a link saying For details and a short historic introduction to MP2. Surely because it is related to MP2 then it should be in this article and not in the MP3 article. Matthuxtable 16:29, 15 March 2006 (UTC)

P.s. There's a section on my talk page if you want me to see your reply as I may forget to check back here. See User talk Matthuxtable Special Reply Section. Many thanks Matthuxtable 16:30, 15 March 2006 (UTC)

How about licensing and patent issues? Are they the same as for MP3?

I heard a rumor that the patents for MP2 are expired. True? The "For details and a short historic introduction to MP2, see MP3." is quite bad. — Omegatron 16:29, 20 July 2006 (UTC)

[edit] Too technical

The second section is just swirling with stuff that just goes right over my head and I consider myself pretty tech-savvy. I think it ought to either be removed or heavily rewritten to follow the KISS principle. — User:ACupOfCoffee@ 17:28, 3 October 2006 (UTC)

I added links for sampling rate and bitrate to the second section. Does that help? Daniel.Cardenas 20:10, 3 October 2006 (UTC)