The Right Thing To Do
I've been lurking on some of the Internet newsgroups recently. It's an interesting experience: If you want to experience audio life as it might be if Stereophile did not exist, you should subscribe to rec.audio.high-end. Despite the newsgroup's title, some of its most vocal denizens seem to have both little interest in high-end audio and much faith in their own belief systems. If I were as sure about one thing as these cyberspace cowboys appear to be about everything, my life would be much tidier. It would also be a lot less rich.
Take digital. If you hang out on r.a.h-e much, you'll read ad nauseam that 16-bit digital is good enough for everyone and that anyone who dares say otherwise, or who—Saints Nyquist and Shannon forbid—prefers analog, is suffering from delusions. "Put it to a double-blind ABX test!" goes up the cry whenever some newbie dares to criticize the sound of CDs and ascribe it to anything other than poor microphones, a poor recording venue, or a hamfisted engineer.
One poster, Siegfried Duraybito, has been flamed so many times his epidermis must be done to a crisp (else he spreads his screen with 45-strength sunblock). Yet I continue to be surprised by my experiences with digital: Either I hear what I don't expect, or I don't hear what I do expect.
Let me tell you two tales:
• I recently uploaded all seven hours of the 20-bit, 4-channel Nagra-D session tapes for Stereophile's forthcoming Robert Silverman Liszt CD (footnote 1) into my Sonic Solutions hard-disk system. (I wanted to start work on the editing during what passes for my leisure time.) I started listening to Liebestraum. Major disappointment. I had remembered the piano as sounding almost luminous: richly reverberant yet subtly detailed. But the recording was more "clangy" than I remembered; the piano sounded smaller, the acoustic drier, the reverberation tails less lusciously long. Had I misremembered? Were my expectations too high? I powered down the system, had a cognac, went to bed, and slept fretfully.
Around 4am I sat bolt upright in bed, went straight to the listening room, and checked the Sonic Solutions audio input/output parameter settings. Input word length was set to 20 bits; output word length was set to...16 bits. I was inadvertently chopping off the four LSBs of my beautiful data! I reset the output word length—that was more like it. Without the truncation, the piano sounded like the 9' New York Steinway it was, and not like some fiberboard knockoff. And the reverberation? Even the sound of my voice on the cue speaker could be heard to light up the church acoustic in a most realistic way.
Wes Phillips came over to do some A/B tests. He heard what I heard—and this was against a background of computer-fan and hard-disk noise. So the next time someone tells you 16-bit digital is good enough, ask them what color the sky is in their world. Or at least ask whether they've actually done any comparisons like this to make them so sure of their philosophical ground.
• At the recent 1996 Winter CES, I made sure to visit recording engineer and erstwhile specialty audio retailer Peter McGrath. Peter was demonstrating the excellent Egglestonworks Andra speakers with Levinson electronics and, among other sources, a Nagra master tape of pianist Valentina Lisiytsa performing virtuoso arrangements by Liszt and Godowsky (Audiofon CD 7205). The sound was to die for. Luminous. Powerful. Like a real piano. I then heard a CD-R cut from the Nagra tape, noise-shaped down to 16-bit data using the Meridian 518. Close, very close. Then Peter played a test pressing whose CD glass master had apparently been cut from a double-speed CD-R.
My jaw dropped. While still a good piano sound in absolute terms, the CD sounded more clangy, the reverberation tails less lusciously long. The pressing plant had assured PM that the data were identical, yet the quality difference was so large that Peter and his associate could identify it 10 times out of 10 in a blind test.
Peter asked me to take both discs home from the CES and do a bit-for-bit file comparison on the Sonic Solutions. To my surprise, the data were identical. Unlike my experience with the first mastering of Stereophile's Concert CD, when a careless transfer had changed the data (see February 1995, p.3), the Lisiytsa CD pressing was identical to the CD-R made directly from the master tape. Yet compared, say, with the differences between two good solid-state amplifiers, the sonic difference between the two discs was enormous. It could only be due to differing levels of word-clock jitter cut right into the disc. If I find that idea alarming—I had always assumed that all data-storage media act as brick walls to upstream jitter—how will the more obstreperous inhabitants of r.a.h-e react to it?
I offer this story to accompany Bob Katz's and Michael Fremer's discussions in this issue on the perils of mastering CDs from jittery double-speed transfer media. As Bob mentions in his letter, Sony Classical's David Smith has developed a system in which a CD's worth of data can be clocked out of RAM to feed the glass-master cutting system. This seems a brute-force way to solve the problem, but if that's what it takes, it must be the right thing to do.
But only if you accept that not all is known regarding digital technology. If you insist that everything is already good enough, then how can it be made better? And "better" is what we're asking for.
Footnote 1: Due for release at HI-FI '96, to be held at Manhattan's Waldorf=Astoria Hotel, May 31 through June 2, where Robert Silverman will be performing live.