"CD Quality": Where Did the Music Go?
I was allowing listeners to compare the 24-bit/88.2kHz-sampled master files for some of my Stereophile recordings with "Red Book" (CD) versions and MP3s encoded at various bit rates. No one had a problem hearing differences between the different versions of these recordings. Yes, I was using a reasonably high-end system assembled by Roy Gregory, editor of UK magazine HiFi+ (see ), but the speakers, two-way Avalon Evolution NP2s, weren't all that expensive at $1995/pair. So when I read in the mainstream media that MP3s are "indistinguishable" from CDs, or that their sound quality is "good enough," I wonder whether the journalists have bothered to do such a comparison. Or were they just parroting what they'd been fed by some PR flack? Or was it just that they'd never been shown what to listen for? (footnote 1)
Not that the degradation is particularly subtle. Stephen Mejias was visiting one morning in early December, to help me pack up the humongous KEF Reference 207/2 speakers that I review in this issue. I took the opportunity to play for him not only some of the recordings mentioned in that review, but also one of the comparisons I'd performed at RMAF. I first played him the 24-bit/88.2kHz file of Cantus performing a new work by Maura Bosch, The Turning, a poignant piece for male choir that will be released on CD in June. Then I played him a 44.1kHz MP3 encoded from the master at 64kbps—about the best "CD quality" you can get from satellite radio. Then I played him the master again.
Stephen rolled his eyes. "The MP3 sounded flat, dull, and lifeless in comparison," he wrote in a subsequent e-mail. "The sense of air and space was lost, and, consequently, the emotional impact was drowned. Blecch. Even the performance suffered—the singers sounded as if they had become tired and sloppy."
This was an insightful comment. As classical recording engineer Tony Faulkner wrote in the December 2007 issue of HiFi Critic, not only do lossy-compression algorithms such as MP3 discard frequency information, they smear the timing of transient information. To perform its analysis, the MP3 encoder has to split the continuous waveform into discrete chunks. This means that any musical attack no longer happens exactly on the beat, but starts just before and ends just after the beat. This effect gets increasingly worse as the bit rate drops, with the result that the musical performance is degraded. "Tired and sloppy," as Stephen put it.
Again, this isn't necessarily subtle. In the comparison I played Stephen, there is a section where the end of one phrase ends with a word ending in s and the next phrase starts with a word beginning in s. With the hi-rez original, you hear a very brief amount of ambient decay between the end of the first sibilant and the beginning of the new one. On the 64kbps MP3, the two s sounds merge into one another. Doubling the bit rate to the 128kbps typical of an iTunes download gives an audible improvement, and increasing it to the 320kbps used for Deutsche Grammophon's new classical downloads restores a reasonable facsimile of the sense that these are real singers singing in a real acoustic—though the difference between that and the hi-rez original is still audible.
These thoughts were triggered by an article written by New York Times opera critic Anthony Tommasini, "Hard Being an Audiophile in an iPod World." Tommasini's thesis was that "over the last decade the ranks of true audiophiles have been thinning, in large part because of the growing popularity of MP3 players and iPods." He does point out that MP3 lack the ability to equal the sound quality of a CD of the same recording, but to him, that's irrelevant: "Convenience was the appeal, and the sound was, well, good enough."
Michael Fremer responds to Tommasini's piece, and to a similar one by the Wall Street Journal's Terry Teachout, in this issue's "Analog Corner" (pp.24–33); and Stereophile's Fred Kaplan wrote an excellent, upbeat essay in response to both for online magazine Slate ("In Defense of Audiophiles"). But it's beyond me how anyone, let alone someone as musically literate as Anthony Tommasini, can feel that the sound quality of a low-bit-rate MP3 is "good enough." In the words of producer Elliot Mazer (Neil Young's Harvest, etc.), in a recent e-mail, "Those of us that have had the pleasure to hear wonderful music played back wonderfully know how good things can be. We must spread the word." These days, when I read the words "CD quality" in the mainstream press in connection with MP3 or HD Radio or satellite radio, I'm beginning to feel like sitting the benighted scribe down in my listening chair and playing him what I played Stephen. Then he'll know what he's been missing!
In the "Loudness Wars," of course, when it comes to the audible degradation of recordings, audiophile sensibilities have long been collateral damage. The higher the average level of a recording—achieved by squashing the transient peaks with a compressor—the louder it sounds, and the record industry seems populated by peculiar people who naïvely believe that more loudness is always more better. Ultimately, you can achieve maximal loudness only by minimizing the music, with the result that such pathologically compressed recordings as Californication, by the Red Hot Chili Peppers (talented musicians all), are amusical dreck. I've been writing about the "Loudness Wars" since December 1999, when, disturbed by the unlistenable sound of Carlos Santana's Supernatural, I took a look at its dynamic range: no more than 8dB. Blecch. My most recent report from the front was my September 2007 article on the making of Attention Screen's Live at Merkin Hall CD, in which my goal was to preserve the full dynamic range of an electric jazz group.
Engineer Bob Katz examined this problem at length in his superb book Mastering Audio: The Art and the Science (Focal Press, 2002; ISBN 0-240-80545-3). In October 2007, in the introduction to a workshop on the subject he organized at the 123rd Audio Engineering Society Convention, renowned mastering engineer Bob Ludwig of Gateway Mastering said that "mastering hot levels has been with us since the LP and 45 singles [but] it has gotten out of hand for CDs and downloads." "From the mastering trenches, I can report that there is a slight shift towards less loud (it could hardly have gone much louder!)," reported Bob in an email. He alerted me to a new music-industry organization, Turn Me Up!, started by engineer Charles Dye, that is campaigning to restore to artists the choice to release more dynamic-sounding records. The name comes from the fact that if the record has a wider dynamic range, it must have a lower average level, though the consumer won't necessarily know that—all he has to do to compensate is turn up the volume.
The Turn Me Up! website includes a two-minute QuickTime tutorial video by Matt Mayfield, as well as a growing list of links to articles on the subject. But between the analog compression being campaigned against by Bobs Ludwig and Katz and the lossy digital compression that is ubiquitous these days, it's hard for those of us who were around at the launch of the CD to accept that, a quarter century later, "CD quality" might be an unattainable goal.
Footnote 1: The Audio Engineering Society has made a CD-ROM available, Perceptual Audio Coders: What to Listen For, to demonstrate the typical degradations in sound quality that result from lossy compression. It costs $20. See www.aes.org/publications/AudioCoding.cfm.