   
|
||
 February 1, 2005 So What is MP3, Anyway? You can’t spend much time reading the audio press without bumping into MP3. I suspect, however, that few of even its greatest fans know what the format actually is.MP3 is a technique for compressing audio files so that they take up less space in a computer’s memory and can be downloaded from the Internet more quickly than a larger, uncompressed file. In audio and video, the process is called perceptual coding or data reduction or (not very accurately) digital compression. The systems used to do this are called coding/decoding schemes, or codecs. MP3 is one of several codecs now in use. Digital recording and the transmission of high-quality audio signals have been with us for more than 30 years. In the early days, the major drawback of digital recording and playback was that they required enormous amounts of data. To record or play back a digital file in real time required very wide bandwidth -- on the order of 1.5 megahertz for a two-channel signal. In codec-speak, that translates to a bit rate of 1500 kilobits per second (1500kbps). This was no problem with such capacious media as optical discs (CDs and LDs). But media with less bandwidth -- digital radio, for instance, or Internet connections -- and/or signals that included more channels, required data compression. Psychoacousticians have long known that the human hearing system reacts to only a relatively small portion of a complex audio waveform; many sounds are masked by other sounds. Empirical evidence of this was obtained by switching on and off the masked sounds in the presence of the masking ones. In many cases the effect was inaudible, which allowed scientists to figure out what was happening and to predict how a particular group of sounds would behave. Their task was to automate that process -- to come up with a circuit that would analyze an audio waveform and remove the parts that wouldn’t be heard anyway. Early efforts in this direction were undertaken by the Motion Pictures Experts Group (MPEG), who was set the task of developing a codec to be used in digital radio broadcasts. The result, MPEG Audio Layer I, involved a compression ratio of 6:1, which meant an overall bit rate of 256kbps for a two-channel signal. But because codecs are usually specified "per channel," MPEG I is known as a 128kbps system. Work on that sort of coding was confined to the lab until 1990, when Philips dropped a bombshell on the audio world with the unveiling of the Digital Compact Cassette (DCC). The format went nowhere, but the most notable thing about it was that it used a form of the MPEG I codec with 4:1 compression, or a bit rate of 192kbps, to squeeze a digital audio signal onto what was, essentially, an ordinary audio cassette. Most audio journalists of my acquaintance were aware of the theory of masking, but few believed that the process could be completely transparent. A series of intensive listening sessions by Philips and others convinced virtually everyone that their codec produced no audible artifacts. The early success of MPEG I made perceptual coding an accepted and permanent part of the audio landscape. Dolby Laboratories came up with AC-3, the algorithm at the core of Dolby Digital surround sound for both theaters and home systems, and Sony developed ATRAC for the MiniDisc. The MPEG people quickly came up with an enhanced codec called MPEG Audio Layer II, and then MPEG-2 Audio, which is used in some DVD and satellite broadcast systems. The most notable enhancement, however, was the MPEG Audio Layer III codec, aka MP3, which was specifically developed to create audio files for computers. The earlier codecs were meant for real-time playing; MP3’s main purpose was speed of downloading. Most codecs are actually families of codecs whose members are able to operate at various bit rates. Generally, it’s accepted that the higher the bit rate, the better the sound, but the urge to save space and time has been such that lower and lower rates have appeared over the years. To find out what that means sonically, the Canadian Communications Research Centre conducted a series of listening tests not long after the appearance of MP3. Some six codec families were included, some of which included as many as four different bit rates. Two different versions (hardware- and software-based) of MPEG II were included, as well as MPEG’s Advanced Audio Coding (or AAC), Dolby’s AC-3, Lucent Technologies’ PAC, and MP3. Several dozen listeners took part in the tests, and when the results were analyzed, the codecs were ranked according to how often the listeners could hear distortions. The clear winners were the top AAC and AC-3 codecs. MP3 ended up in the middle, its single bit rate of 128kbps slightly outperforming both of its MPEG II stablemates at 160kbps. But PAC, at 96kbps, was almost as good, and AAC, at 128kbps, was far better. After that, the codecs were improved, and several years later another set of listening tests was conducted, by Sound & Vision magazine, in New York City. The systems compared were MP3, AAC, and Microsoft’s Windows Media Audio (WMA). The listeners were techies from the MPEG group, Dolby Laboratories, and Microsoft. As in the earlier CCRC tests, AAC came out on top, followed by WMA. Trailing the field was MP3, which prompted Sound & Vision to write, "Makers of compressed-data playback devices . . . would be doing critical listeners a service by incorporating [AAC and WMA] into their products." All three codecs are now reasonably common, and MP3 now offers higher bit rates, which improve its performance considerably. But although Apple’s immensely popular iPod player supports MP3 at several bit rates, Apple’s own iTunes music-download service offers only AAC files. ...Ian G. Masters
|
||
|
