Listen Up
Better than sex? That's the rap MP3 is getting on the Internet. We decode the digital audio universe.
Brian Dipert, Contributing Editor

I mean, we're talking about sex here. The primal urge. The industry whose revenue stream dominated the early days of the Internet, much as it single-handedly powered the fledgling videocassette and video CD businesses. The commodity that first drove a number of innovations now considered (for better or worse) mainstream; streaming video, Netcams, chat, secure credit card transactions, spam email, cookies and banner ads.

When something topples the almighty sex from its top search-engine spot, it's time for the digerati to sit up and take notice. So listen up, gentle readers, as I unravel the mysteries of digital audio creation, distribution, and playback. In addition to explaining what digital audio is, it's important to explain why the phenomenon has happened now, versus in the past, and by extrapolation make some predictions as to where the technology may be headed. So I shall.

Stop, hey, what's that sound?

Let's begin with a little math. High-fidelity compact disc audio uses a 44.1-kHz sampling rate, with each sample 16 bits in size, and dual (right and left) audio channels. Punch the numbers into your calculator and you'll find that these specifications translate to just over 1.4 Mbits/sec, or 175,000 Kbytes/sec, of required storage capacity and transmission bandwidth. That's a big WAV file; more than 10.5 Mbytes per minute of music. And don't assume that lossless compression techniques like those found in PKZip will by themselves put that bitstream on much of a diet. The data's just too random. Next-generation DVD Audio supports up to eight audio channels, at sampling rates up to 96 kHz (multi-channel) and 192 kHz (dual-channel), and with per-channel sample sizes up to 24 bits. Even after Meridian Lossless Packing squeezes everything down, that's a whole lot of bits, folks.

OK, now consider everyone's favorite lossy compression standard of the moment, MP3. Shorthand for the layer 3 audio compression format built into the MPEG-1 standard, and therefore also into MPEG-2 along with multi-channel Dolby Digital AC-3 and AAC (Advanced Audio Coding), MP3 is one of several perceptual audio encoding schemes currently vying for consumer, and therefore content-developer, mindshare. Perceptual audio compression, simply stated, tosses out information that, even if it existed, our auditory system wouldn't be able to process.

For example, a sound of a given frequency will tend to mask out information in nearby frequencies of even slightly less intensity. Psycho-acoustic transfer functions also take advantage of human beings' absolute hearing thresholds. Lossy compression techniques incorporate plenty of lossless compression tricks, too, such as run-length encoding, differencing, statistical coding and substitutional compression. And they focus much of their efforts on the higher-frequency bands, because the information content here naturally has greater sample-to-sample variation than, say, in the deep bass range.

128-kbit/sec-encoded MP3, the most common bitstream rate, takes up less than a tenth the storage space of 44.1-kHz-sampled stereo CD audio. That's a big deal when you're talking about flash memory cards at $4 or so per megabyte, or if you're using an old PC with a tiny hard drive. Especially if you're a budget-constrained high-school or college student (who, coincidentally, doesn't care to understand the nuances of the Home Recording Act of 1992). It's also a big deal when you're trying to use a narrow straw (your 56-kbit/sec analog modem) to suck down a five-minute song, a 50-minute album or a five-hour concert. All other factors being equal or nearly so, even we ADSL and cable-modem users appreciate a 12-times-faster download rate.

MP3 has garnered the majority of consumer attention to date simply because the encoding and decoding software are free, easily obtainable, and easy to use (see sidebar, "Resources"). Little Johnny can "rip" (extract to digital audio files) his CD collection with a few mouse clicks and without shelling out a dime beyond the PC his parents already gave him. And yes, he can share these files with his friends, and listen to their stuff, too, but we'll talk more about digital audio security in a future article and also in the sidebar, "The record industry versus the human brain." Some folks think that the latest generation computer CD-ROM drives with digital audio extraction (DAE) capability are partly to blame (or thank, depending on your perspective) for the exploding popularity of digital audio. I agree with them, but not for the most commonly touted reasons.

The pristine quality of digital-to-digital transfer that's got record execs so freaked out is a secondary benefit to DAE, for the same reason that 128-kbit/sec MP3 audio is good enough for most folks. A first-generation digital-to-analog conversion coming from the CD-ROM drive to the sound card, followed by an analog-to-digital reconversion in the encoder, still produces darn-good-sounding digital audio, unless you're using a PC with really cheap D/A and A/D converters, or one that generates an obscene amount of EMI (electromagnetic interference). But here's the catch: with a DAE-enabled system, you can rip a bitstream as fast as your CD-ROM drive can spin, instead of being held back by the 1X audio CD transfer rate.

MUSICAL INSTRUMENTS: Creative's Nomad player and some of the many freely downloadable software tools for manipulating digital audio.

Now for the formats (see sidebar, "Resources"). MP3 we've already mentioned. Two other contenders, Liquid Audio and Lucent Technology's Perceptual Audio Coder (ePAC), have been around roughly as long as MP3 but haven't achieved a comparable level of popularity. The proprietary angle to the Liquid Audio Player isn't the codec, it's the company's watermarking scheme. Liquid Audio uses both MPEG-2's industry-standardized AC-3 and AAC. AAC's claim to fame, like that of ePAC, is CD-quality audio at lower bit rates (roughly 96 kbits/sec) than MP3. Ironically, though, AT&T spinoff a2b music currently uses a watermarked version of AAC, not ePAC. Lucent plans, however, to introduce a line of portable digital audio players based on ePAC in the near future.

RealNetworks' RealAudio format also includes a 96-kbit/sec claimed CD-quality option, but today RealAudio is mostly used for low-quality streaming music previews. And no discussion of digital audio formats is complete without mention of the 10-ton gorilla of the computer industry, Microsoft. The company's Windows Media (WMA) format is a relative latecomer to the scene, but it's made impressive gains.

Some of this success is certainly due to Microsoft's deep pockets, which have resulted in a number of well-known artists distributing their music in its format (support that MP3 and Liquid Audio, and to a far greater extent ePAC, lack). Some of this success is also due to the fact that Microsoft freely distributes its encoder, decoder and rights-management software, leading to their integration into almost all of today's top digital audio creation toolsets. The key point to WMA's popularity, though, is its very high quality at low bit rates. Reviewers consistently miss this nuance when they compare various audio formats at equivalent bit rate settings. Microsoft claims that 64-kbit/sec WMA sounds as good as 128-kbit/sec MP3, if not better. From personal experience, I have to say I agree with them.

Format grab-bag

Several other contenders bear at least brief mention. IBM is said to be developing its own proprietary compression formats as part of its broader-vision Electronic Media Management System, code-named the Madison Project. Sony's first stabs at portable digital audio players, the Memory Stick Walkman and Music Clip introduced at November 1999's Comdex, will use the MiniDisc-derived ATRAC (Adaptive Transform Acoustic Coupling) format. Apple has built support in QuickTime for QDesign's audio codec, which is particularly adept at handling sub-56-kbit/sec analog modem bit rates.

An NTT-developed and Yamaha-marketed format variously called TwinVQ (Transform-Domain Weighted Interleave Vector Quantization) or VQF claims to deliver equivalent quality to MP3 with file sizes at least 30% smaller. VQF's encoding and decoding algorithms are, however, more processing-intensive than those of MP3 (which, like other pieces of MPEG, is an asymmetrical approach with comparatively complex encoding and simple decoding). And the MPEG committee is rolling out the MPEG-4 audio compression standard, based on VQF for low-bit rate applications and AAC for high-bit rate environments.

Live music afficianados claim, with some justification, that aggressive lossy compression does a particularly bad job of handling concert-recorded material. The reasons here are two-fold: live music dynamic range and sample-to-sample amplitude variations tend to exceed those of most studio material, and the less-than-ideal acoustic environment injects a lot of high-frequency noise into the recording. Variable-bit-rate MP3 would help minimize these shortcomings, but not all encoders and transcoders support this MP3 option, nor do all players. Softsound's lossless Shorten format, which on average will compress a WAV file by 50% or so and decompresses it to its original quality, is a popular alternative to MP3. Other compression options claiming lossless performance include MUSICompress and WAVPACK. Softsound also offers the ATELP lossy coder, which targets high-quality audio at 73 kbits/sec.

Bye, bye, brick-and-mortar

Now that you know what digital audio is, where can you get it? Specifically, how can you find music you don't already own? At a growing number of popular sites (see sidebar, "Resources"). Their business models vary widely; in some cases they extract a portion of the revenue from each sale (though, typically, a smaller cut than a record company would grab), and some take the next step of charging artists to encode and upload the track(s) to their servers. Others, such as MP3.com, don't charge the artists for posting or downloads. Instead, they make most of their money from banner advertising and, to a lesser extent, by burning tracks onto CD-Rs and selling them to interested listeners. Ironically, MP3.com's name is less than accurate; the company has MP3 roots but at this point is format-agnostic.

A widespread underground network, consisting primarily of college students using their university-allocated server space and bandwidth, as well as hobbyists who have converted an old PC and a cable modem or DSL connection into their very own FTP server, also exists to distribute digital audio recordings. Use a search engine such as Lycos' MP3 Search to find the music you're looking for, but don't be surprised if you get a lot of "file not found" messages, especially if you're looking for things you shouldn't (ie, copyright-protected songs). FTP sites and Internet newsgroups are, however, ideal places to find recordings from acts such as the Dave Matthews Band, the Grateful Dead and spinoffs, and Phish, all of whom allow recording and free redistribution of their live concerts.

The record companies aren't letting themselves be written out of the picture, though you're more likely to find short-duration, low-quality previews of songs than full-length high fidelity clips on their sites. Record store chains such as Tower Records are getting into the act, too. And, to one degree or another, many artists are themselves embracing digital media (usually out of frustration with the record labels). The Artist (Formerly Known as Prince) originally envisioned bypassing record companies and selling his music directly to the public through his site, though he's recently backed off somewhat from this stance. More commonly, you'll find either time-limited songs or unlimited-play alternate versions of tracks found on the CD, plus video clips. Hard-rock band Metallica recently took the avant-garde step of making available, in streaming Windows Media format, all the tracks from its S&M album, starting four days before the disc showed up in stores.

Play that funky music, new toy

The first few generations of stand-alone digital audio players have been mostly single-format MP3 devices. Diamond Multimedia's Rio PMP300 led the charge, partially because it was first-to-market and partially because other manufacturers sat on the sidelines waiting to see if Diamond would survive the record-industry lawsuits that followed Rio's launch. I previewed several MP3 players from other companies behind closed doors at January 1999's Consumer Electronics Show, players that only recently appeared in stores.

Today's MP3-only devices need only incorporate a hard-coded MP3 processing engine, such as one of Micronas' chips. Future-generation players, on the other hand, may have to support multiple audio formats (though alternatively this task could be handled by a media-management program running on the host computer) and will probably also comprehend secure transactions, all of which implies a more flexible, more powerful processor. Predictably, the major DSP manufacturers (Analog Devices, Lucent, Motorola, Philips, and Texas Instruments) are all rushing cost- and power-optimized architectures to market, and Cirrus Logic is also aggressively promoting a family of multiformat ARM7-based digital audio DSPs.

What if you'd like to play digital audio files on your computer? The leading media players, Apple's QuickTime, Microsoft's Media Player, and Real Networks' RealPlayer, all support MP3, and RealPlayer adds support for a2b music and Liquid Audio formats (both of these companies also offer free stand-alone players) and ePAC. Other well-known freeware players include Nullsoft's WinAmp and the Sonique Player. Both Sonique and Winamp let you export audio as a WAV file for subsequent burning onto a CD, and WinAmp also inputs and outputs Windows Media format.

And if you'd like to transfer music from your CDs to MP3, for playback in your shiny new portable MP3 player Christmas present? For this task you'll need several pieces of software; one to access the CD-ROM drive and "rip" the data to a WAV file, another to handle the WAV-to-MP3 transcoding, and a third to manage the flash memory inside the player and transfer new files to it. Fortunately, these pieces are merging in the latest versions of audio media-management software—such as MusicMatch's Jukebox, Real Networks' RealJukebox, and RioPort's Audio Manager.

I'm not able to reliably operate the first two programs on one of my desktop PCs, but they run like a charm on my newer machine, and RioPort Audio Manager hums along without problems on both computers. MusicMatch Jukebox works fine on the first PC until I attempt to "rip" CAB-encoded information off an audio CD; with already-existing WAV files it transcodes without crashing. I've gotten around the problem by installing an upgraded (although unsupported) version of Windows 9x file CDFS.VXD. Now, when I view the contents of the audio CD, I see not only the CAB files but also WAV files corresponding to the audio CD's songs, in mono and stereo, 8 and 16 bit, and at 11.025-, 22.05-, and 44.1-kHz sampling-rate versions. CDFS.VXD also eliminates the time- and HDD space-consuming interim step of converting the streaming CD audio data to a WAV file before transcoding it to MP3. Andre Wiethoff's Exact Audio Copy is another highly regarded, very accurate (and free) audio "ripper," and Tord Jansson's BladeEnc is a well-known freeware WAV-to-MP3 encoder.

The free versions of MusicMatch Jukebox and RealNetworks' RealJukebox only encode MP3 files to less-than-CD 96-kbit/sec quality. RioPort Audio Manager's free version limits the number of files you can MP3-convert. A diverse set of perfectly capable freeware alternatives exist, though, and I have a funny feeling there's likely to be a Microsoft-developed alternative coming along eventually, too. I suspect, therefore, that at some point all the "jukebox" manufacturers will switch to a Netscape-like business model that doesn't depend on software sales but generates percentage revenue with every audio track purchased through their software portals. CD-to-MP3 extraction alternatives include Xing Technology's Audio Catalyst, Jackie Franck's Audiograbber, Markus Barth's CD Copy, Christoph Schmelnik's Digital Audio Copy and Jukka Poikolainen's Easy CDDA Extractor. And there's plenty more: Check out the software section on MP3.com.

Finally, what if you're a live-concert attendee like me, and you'd like to capture a digital audio memory of a show? First, make sure that the band you're going to see allows taping; most don't, and in this case, your microphones and tape deck probably will be confiscated and you will be asked to leave. Most PCs' sound cards or motherboard-resident sound chips provide line inputs that you can attach to the line output of any tape deck. If you've recorded the music with a portable digital audio tape (DAT) deck, and you want to do a direct digital transfer to your hard drive, you're restricted to a shorter list of soundcards—ones that include SPDIF digital inputs. Creative Labs' Sound Blaster Live series or Turtle Beach's Montego II are examples. Keep in mind, though, that some of these cards will hardware-upsample all inputs to 48 kHz, so you'll have to subsequently convert the sampling rate to a more CD-friendly 44.1-kHz before burning a disc.

Alternatively, you can run the DAT deck's digital outputs through a SPDIF-to-USB conversion box such as Ego Systems' U24, Opcode Systems' DATPort or SonicPort, or Roland's UA-30 AudioCanvas. Higher-powered descendents of Microsoft's Sound Recorder such as GoldWave, Sonic Foundry's Sound Forge and Syntrillium's Cool Edit take it from there, pulling data off the analog line or digital USB inputs, offering you a variety of editing and processing functions, and saving your music in a variety of digital audio formats. To burn your digital audio files onto a CD-R or CD-RW disc, turn to software such as Adaptec's Easy CD Creator, Ahead Software's Nero, or Sonic Foundry's Siren Jukebox. CV

Author information

Contributing Editor and Phish fan Brian Dipert is a technical editor for EDN (www.ednmag.com). He assures us that his hard disk harbors no files that might be construed as violating copyright laws.