May We Serve You?

There once was a time when audio/video components didn't have USB or Ethernet ports. When "kilobits per second" was not a hi-fi term. When a kid who stumbled over a stack of LPs in a dumpster actually knew what they were.

Welcome to the "00s," or whatever we're calling this decade in which the centerpiece of a high-performance home entertainment system may well come to be, uh . . . what do we call these things? Personal audio recorders? Hard-disk players? MP3 hosts? How about music servers?

Whatever you call 'em, we rounded up five models - Hewlett-Packard's de100c ($1,000), the Imerge SoundServer S1000 ($1,500), Onkyo's MB-S1 Music Library ($800), the Rio Central from SonicBlue ($1,500), and ZapMedia's ZapStation ($1,500) - for hands-on evaluation and lab tests. And you can bet there'll be more. Computer companies and traditional A/V companies alike are scrambling to meet the challenge of integrating entertainment options into our increasingly digital homes, all hoping to deliver the killer appliance that will merge the computer and A/V worlds once and for all.

Though our five music servers are impressively varied, they do have some features in common. They all read standard music CDs and convert, or "transcode," the uncompressed (PCM format) digital audio data into a compressed format that's stored on a large internal hard-disk drive with a capacity of 20 to 40 gigabytes (GB). A 40-GB drive can hold upwards of 650 CDs' worth of compressed music. Equally important, all of that music, which you can search and sort through with ease, is accessible almost instantaneously.

Better still, each stored track can incorporate text data identifying the artist, the album it came from, the musical genre, and so on. Without such information, a huge digital jukebox would be pretty useless - how would you ever find what you wanted to listen to? But having text IDs also lets you create custom playlists. You might hand select a "Dance Party" playlist, for example, or use automated playlist creation to bring together all your piano-concerto recordings (sorted by title) or all tracks by Miles Davis (sorted by artist). Of course, manually entering the track title, album title, genre, and artist for thousands of selections using buttons on a remote control or front panel might get a little tiresome. That's why all but one of these servers can access Gracenote's Web-based CDDB Music Recognition Service to identify discs and tracks and to download text information.

Four of these servers can play CDs the old-fashioned way, too - one even plays DVDs. A couple are CD recorders as well, able to burn a disc in uncompressed PCM or compressed MP3 format. And all include a USB port so you can transfer audio files between the server and a Wintel PC or a portable player, though in one case (Imerge) these features aren't yet supported.

How Do They Sound? Before we go any further, you probably want to know how these things sound. In a nutshell, compressed audio - whether it's in MP3, Windows Media Audio (WMA), or any other format - can never sound exactly the same as the original, uncompressed CD. That said, it can come extraordinarily close.

Of course, digital audio data can be compressed to various degrees, ranging from about 4:1 to as much as 50:1, in a tradeoff between storage efficiency and sound quality. At the highest compression - or lowest bit rate (the amount of data allocated per second of audio) - the sound is worse than AM radio. With moderate to low compression, however, the sound can be practically indistinguishable from the original, especially for casual rather than critical listening.

I listened to MP3 tracks ripped by each server at its highest (least compressed) bit rate - to see if it was capable of storing compressed audio that's virtually indistinguishable from the original - and also at 128 kilobits per second (kbps), which is as close to a de facto bit-rate standard for compressed audio as we have. (Conveniently, 128 kbps increases storage efficiency over uncompressed CD audio by about a factor of ten.) In each case, I dubbed some or all of several CDs that I know extremely well, including Natalie Merchant's Motherland, Schumann's Davidsbündlertänze, a demanding set of piano pieces (performed by Andreas Haefliger), Steely Dan's Two Against Nature, and Amanda McBroom's impeccably recorded Growing Up in Hollywood (an old Sheffield Labs disc).

The results were pretty much as I'd have predicted. Even at the fairly severe 128-kbps rate, each of the five servers sounded surprisingly good. On the other hand, none sounded perfect. Sound quality at 128 kbps nearly matched the original CDs, with only occasional telltale compression artifacts - like a "smooshy" snare-drum hit or a faint veil of fizzy, swirly noise - that were discernible only in highly critical listening, mostly done with headphones. Similarly, at each server's maximum bit rate, I couldn't find any musical passages betraying that what I was hearing from the hard drive was compressed rather than uncompressed.

The bottom line: You can use any of these servers for serious music listening. But I recommend choosing bit rates on an album-by-album basis. I wouldn't squander 320 kbps on, say, Paul Revere and the Raiders' Greatest Hits - 160 kbps, 128 kbps, or perhaps even less will do just fine for that. But I'd probably use the highest rate for a meticulously produced, contemporary recording like Acoustic Soul by India.Arie - or almost any decent harpsichord recording, if you like that sort of thing, since harpsichord is the most difficult instrument for MP3 codecs to handle.

Features Comparison In The Lab