17.9 MP3

17.9 MP3

We get so many questions about MP3 that we felt compelled to explain briefly what MP3 is and how to use it. Here seems as good a place as any. MP3 is shorthand for MPEG-1 Layer 3 audio compression. MP3 has quickly become the most popular format for storing audio on computers. Although the music industry, concerned about piracy, has done its best to kill MP3 by filing lawsuits and introducing competing secure digital music formats, MP3 thrives and is likely to continue doing so.

The raison d'être for MP3 is that it compresses audio data significantly while maintaining acceptable audio quality. CD-Audio is recorded in stereo at 44.1 kHz with 16-bit sampling, and stored uncompressed as a 150 KB/s bit stream. Storing one minute of CD-Audio therefore requires 9,000 KB. A standard CD stores up to 74 minutes of audio, which translates to about 650 MB. Current hard disks are huge and cheap, but, at about 1.6 hours/GB, storing even a small CD collection on hard disk in CD-Audio format is impractical.

Enter MP3. MP3 uses lossy variable compression, which means that it stores important data—such as foreground sounds in the middle ranges to which the human ear is most sensitive—using moderate compression, but uses high compression for less important data, such as very high and very low tones and quiet background tones that are overlaid by louder foreground tones. Some data is discarded entirely, such as audio that is identical on both stereo channels and audio that is below the human hearing threshold. MP3 allows encoding at different bit rates, which correspond to different overall compression levels. The most commonly used bit rates are:

64 kb/s (19.2X compression): At this bit rate, often called AM radio quality or voice quality, a one-hour CD compresses to only about 30 MB. Sound quality, however, is noticeably sub-par, even when played on inexpensive computer speakers. Use this bit rate only for encoding voice-only audio, for which it is perfectly acceptable.
128 kb/s (9.6X compression): At this bit rate, often inaccurately called FM radio quality, a one-hour CD compresses to about 60 MB, or 1 MB/min. Sound quality may be indistinguishable from a CD when played on inexpensive computer speakers, but the difference is readily apparent with good speakers or headphones. Most pirated music on the Web is encoded at 128 kb/s. We think this bit rate falls unfortunately in the middle—not good enough to be "good," and not small enough to be "small." We would never use it to encode classical music, but many people find it good enough for rock.
256 kb/s (4.8X compression): At this bit rate, which really is FM radio quality, a one-hour CD compresses to about 120 MB. Sound quality is, for many people, nearly indistinguishable from a CD, even when played on good-quality computer speakers or headphones. We recommend this bit rate for all but the most discerning listeners using top-quality computer speakers in a quiet environment
320 kb/s (3.8X compression): At this bit rate, usually called CD quality, a one-hour CD compresses to about 150 MB. Nearly everyone finds 320 kb/s MP3 files effectively indistinguishable from CD audio. Discerning listeners with top-notch equipment can usually discriminate between them, often describing the MP3 audio subjectively as "lacking sparkle in the highs" or something similar. We can't tell the difference, though, and recommend this bit rate for those who listen to MP3s on good home audio equipment.

In addition to these "standard" bit rates, some people striving to optimize audio quality versus storage space use intermediate bit rates, such as 160 kb/s and 192 kb/s. Not all MP3 players support all bit rates, so make sure the player you want to use supports the bit rate at which you encode your CDs. If you're undecided between two bit rates, use the higher. Many people encode their entire CD collection at a lower bit rate only to find that the quality is unacceptable and they have to re-encode at a higher bit rate. Note that there are two types of MP3 bit rates:

Fixed bit rate MP3: Fixed bit rate MP3 keeps the bit rate constant and varies the compression level to fit the content to the available space. The same number of bits per second are stored regardless of what happens to be going on in the music at any given moment. For example, if you use 128 kilobit/sec fixed bit rate MP3, one second of complete silence is stored as 128 kilobits, just as one second of complex chamber music is. Accordingly, the level of compression used is varied to fit the fixed bit rate, which also varies the quality of the stored music dynamically. That is, the one second of silence will be stored with perfect fidelity at 128 Kb/s, while the one second of chamber music will have lower fidelity because a higher degree of compression was needed to fit its data into 128 kilobits.
Variable bit rate MP3: Variable bit rate MP3 keeps the compression level (and therefore sound quality) constant and varies the number of bits stored per second. 128 Kb/s variable bit rate MP3 stores an average of 128 Kb/s, but gives better fidelity than 128 Kb/s fixed bit rate MP3 because variable bit rate MP3 uses the same number of bits more efficiently and effectively. In general, a variable bit rate MP3 recording sounds like a fixed bit rate MP3 recording made at the next higher level. For example, a 128 Kb/s variable bit rate MP3 recording has audio fidelity similar to that of a 256 Kb/s fixed bit rate MP3 recording.

17.9.1 MP3 Software

Transferring music from CD to MP3 files requires software to perform the following functions:

Digital Audio Extraction: The process of copying the original digital audio data from a CD—properly termed Digital Audio Extraction (DAE)—is usually called ripping. The software used to do it is called a ripper. Ripping extracts an exact digital copy of the audio data and stores it as a .wav file on the hard disk. These are the same size as the original data—650 MB or so for a full CD. Rippers vary greatly in speed, features, and interface, but using any competent ripper produces the same result—an exact copy of the audio data in .wav format. Although ripping speed is partially dependent on the ripper, the major determinant is the speed at which the CD-ROM drive performs DAE. Not all CD-ROM drives can perform DAE, and many that do rip at 1X or less. That is, ripping one hour of music may take one hour or more. We use the ripper bundled with our Plextor CD-ROM drives, which typically rip at 14X to 20X and require only a few minutes to rip an entire CD. Numerous rippers, many of them free, are readily available on the Net. (http://software.mp3.com/software)
MP3 encoding: The process of compressing .wav source files and storing them as .mp3 files is called encoding. The software used to encode MP3s is called an encoder or compressor. Encoding is CPU-intensive, so although the specific encoder used and the degree of compression both affect performance, CPU speed is the main determinant. A slow PC may encode at 0.25X, taking four hours to encode one hour of music. A fast PC may encode at 1X or faster. Many encoders allow batch processing .wav files overnight. We use BladeEnc (http://bladeenc.mp3.no), which, although it has a minimalist text-based interface, is free, reasonably fast, supports a broad range of compression levels, and produces clean MP3 files. As with rippers, numerous free or inexpensive encoders are readily available for download.

Although ripping and encoding are distinct processes, some software combine the two. One such product we've used successfully is Media Jukebox (http://www.musicex.com/mediajukebox).

Playing MP3 files requires MP3 player software. The Microsoft Windows Media Player, bundled with recent versions of Internet Explorer and downloadable from http://www.microsoft.com/windows/mediaplayer suffices for casual use. To download a more capable MP3 player, visit http://software.mp3.com/software, where numerous inexpensive or free MP3 players are available. The most popular MP3 player, WinAmp, formerly shareware but now free, is one we've used successfully.

17.9.2 MP3 Hardware

Depending on how you use MP3, you may need little or no additional hardware:

Listening to MP3 files requires nothing more than a sound card and speakers or headphones.
Creating MP3 files from audio CDs requires a CD-ROM drive that supports Digital Audio Extraction. Most CD-ROM drives made after 1998 are DAE-capable, although more recent and better-quality CD-ROM drives provide both faster and cleaner extraction. That last may seem an odd statement. After all, isn't a digital copy always an exact duplicate of the original? Not necessarily. CD-Audio discs lack the second level ECC code that CD-ROM discs include. That means that the copy may differ from the source. A few flipped bits usually cause no audible differences in the copy, but we prefer to extract source files as cleanly as possible. For that, we have found Plextor SCSI CD-ROM drives to be the best available.
Creating CDs from MP3 files requires a CD-R(W) drive, with which you can create two distinct types of CD. First, you can decompress the MP3 files and write them in CD-Audio format to the CD. This results in a normal CD-Audio disc—albeit with degraded sound quality relative to the original CD-Audio source—which is playable in any CD-ROM drive or CD player that supports multisession, which is to say all recent CD-ROM drives and most recent players. Second, you can write the MP3 files directly to CD, which allows creating a CD with four to ten hours of music on it. You can play these MP3 files on any computer, just as you would if they were stored on the hard disk. Although MP3 CDs are not playable in standard CD players, some new home and car CD players can play them.

17.9.3 MP3 Pro

Announced in mid-2001, MP3 Pro is the second generation of MP3. MP3 Pro products have just begun to ship as of late 2001. The promise of MP3 Pro is to provide equivalent sound quality to MP3 with files half the size. For example, a MP3 Pro file that is the same size as a 64 Kb/s MP3 file has sound quality equivalent to a 128 Kb/s MP3 file. Although we have not done any formal analysis, based on our listening tests that claim appears to be true.

MP3 Pro works its magic by splitting the source into two audio streams. The first stream contains low-frequency information, and is compressed using standard MP3 encoding. The second stream contains high-frequency information, and is compressed using the new MP3 Pro algorithm.

When an MP3 Pro file is played with an MP3 Pro compliant decoder, the decoder reassembles these two streams, resulting in full-bandwidth playback and audio quality much higher than standard MP3 provides with the same file size. An MP3 Pro decoder can also play standard MP3 files, although of course without the enhanced audio quality of MP3 Pro. Although a standard MP3 decoder can play MP3 Pro files, it ignores the high-frequency audio stream that uses MP3 Pro encoding, and so plays only the low-frequency stream, which results in degraded audio quality.

As of June 2002, MP3 Pro encoders and players are just starting to become available. This delay is probably attributable to the higher licensing fees charged for MP3 Pro, but nevertheless we expect MP3 Pro to largely replace standard MP3 by the end of 2002. As of now, if you want to encode MP3 Pro, your choices are still relatively limited. The best MP3 Pro encoder we know of is the $19 MP3 Pro plug-in available for Nero Burning ROM, our favorite CD/DVD burning software. If you'd like to try MP3 Pro before you buy it, download the demo version of the Nero encoder from http://www.nero.com/en/download.htm. This version is fully functional, but limited to encoding 30 tracks. If you're not using Nero to burn CDs, download the trial version of Nero Burning ROM while you're at it.