MP3 storage methods
MP3 storage methods
The MP3 standard does not specify any precise mathematical algorithm for encoding, leaving its design entirely to the encoder developers. Instead, it defines a general coding scheme and also the format of the encoded frame. The frame sequences themselves can be streamed (this process is called streaming) or stored in files.
An MP3 file, like a stream, consists of a sequence of frames with random information between them. The basic requirement is that there must be no overlap with the beginning of a frame.
Often, a standard WAV meta-audio format header is added to the frame sequence, resulting in what is known as a WAV-MP3. (A bit more on the latter will be discussed later, when the ACM pro codec is described.) Even more often an ID3v2 information block is added to an MP3 converted file, containing information about the artist, genre, song title, and other such information about the track. It is added to the end of the file. No one has thought of putting anything in the middle yet. Although, generally speaking, it may be interesting to insert a special tag for VBR with information about what part of the track we're actually in.
The nature of lossy encoding
Low bitrates always cut off small, relatively quiet details, the presence or absence of which often seriously changes the emotional coloration of the song, giving or taking away such effects as a sense of 'crystal' clarity (to the extent that it is present in CD Audio). In addition, according to the psychoacoustic model, higher (above 16kHz) frequencies at low bitrates are encoded at a very low priority.
Further, there are different features of the encoders. For example, FhG IIS encoders have 'blurred' upper frequencies at 128kbs, so there is a 'lisp' effect, while ISO-based encoders have 'chime' instead. Most likely, it is due to the different attitude to frequencies above 16kHz of these encoders. In the Lame encoder, by the way, these are cut by default, which increases the encoding quality by 128kbs.
At higher bitrates, following psychoacoustic model developed by FhG IIS, only errors made while writing encoder could cause problems. However, due to not too much concern of FhG IIS about quality of its encoders at high bitrates, it appeared more than once that new version sounds sometimes even a bit worse than old one. I guess the reason lies in insufficient or wrong relaxation of psychoacoustic model limitations when the bitrate is increased. Again, the first serious customer, and the problem disappears.
Which files can be converted to MP3 format?
All encoders support WAV PCM encoding. AIFF is supported by many. Mp3enc can stream raw data without an intermediate file, in PCM format. Both mpegEnc and Cdex offer encoding directly from CD-ROM. Lame is distributed in source code form and can be taught in any format.
What is the status of MP3 encoders?
All Fraunhofer IIS and Xing Tech products are commercial, but full versions can usually be found on FTP search engines.
BladeEnc, mpegEnc are freeware, but FhG IIS wanted to make their development impossible. This is done in a very simple way - you have to pay a fee to FhG IIS for distributing the coders, even if the coder is distributed for free. Therefore, free distribution of mpegEnc is not possible, and difficulties may arise with the distribution of BladeEnc in spite of the special policy of the author's country regarding patents on mathematical algorithms.
Lame is not in danger of such a fate, since it is distributed as a patch to the ISO source code and cannot be recognized as an encoder by itself.