As We See It

"Everybody, including myself, was astonished to find that it was impossible to distinguish between my own voice, and Mr. Edison's re-creation of it."—Anna Case, Metropolitan Opera Soprano, 1915

"It is my sincerest hope that our successors regard today's pronouncements of digital audio sound quality with the same combination of humor and incredulity with which we view Anna Case's assessment of Mr. Edison's machine."—Robert Harley, audio writer, 1990

When I wrote that conclusion to an editorial in the December 1990 issue of Stereophile, I expected that it would be years or even decades before we thought of today's digital audio as being as anachronistic as the Edison cylinder. Digital audio sound quality was so far behind the best analog that I couldn't imagine a scenario in which a digitally encoded audio signal could meet or exceed analog performance in the immediate future. This pessimistic assessment was partially fostered by the grim realization that the world was stuck with a sampling rate of 44.1kHz and 16-bit quantization. Any wholesale improvements in digital audio would necessitate overthrowing the established standard—an unlikely prospect.

An astonishingly short two years after writing that, I can report that digital audio has taken a significant step forward in its inexorable march toward superiority over analog. The development to which I refer is called High Definition (HDCD). This is the digital encoding/decoding system developed by Keith Johnson and Michael Pflaumer that I described in last November's Stereophile. HDCD is a sophisticated analog/digital converter and proprietary encoding process that produces a digital signal compatible with all existing digital formats and hardware—including the Compact Disc. The second part of HDCD is an optional decoder that would be incorporated into CD players and digital processors. Discs encoded with HDCD will still play back on standard hardware without the decoder, and sound better than conventional CDs, but the process's full potential is realized only by playing HDCD-encoded discs through an HDCD decoder. Note that the decoder is not a requirement, but an enhancement. At the time of my first report, I hadn't heard the process and merely reported its existence.

Since that update, however, I've had a chance to hear HDCD firsthand under good conditions through an excellent playback system. During our stay in San Francisco for the Audio Engineering Society convention in October, John Atkinson and I visited Pacific Microsonics (the company formed to develop and market HDCD) for a private demonstration. The playback system consisted of the wonderful Artemis EOS loudspeakers (augmented with a pair of Keith's subwoofers) driven by a Spectral DMA-80 and Keith's own subwoofer amplifier. The source was first-generation master tapes played back on Keith's unique built-from-scratch analog tape machine. The sound was extraordinarily good; hearing Keith's first-generation analog master tapes was a real treat.

The system was set up so that we could listen to the analog tape directly, or after its output had been digitally encoded and decoded by the HDCD system. A DAT machine was in the digital signal path—with no tape in it—to confirm that the digital signal was indeed compatible with the existing CD standard of 16-bit word length and 44.1kHz sampling frequency.

It's one thing to compare an AAA LP with a DDD CD of the same musical event (Reference Recordings releases, for example, which I use in evaluating D/A converters), but far more rigorous to compare a digitized signal with a first-generation analog master tape—particularly Keith's high-resolution recordings. The original analog master is a much higher standard to strive for than an LP. Any weaknesses in the digital system will be more obvious when compared to such a superb source.

After getting familiar with the system by listening only to analog tape, the HDCD encode/decode system was switched into the signal path. I sat there incredulous at what I didn't hear: the signal after HDCD encoding/decoding sounded virtually identical to the first-generation analog master tape. There was an almost complete absence of the familiar digital artifacts: glare, stridency, loss of depth and space, flat sterility, congestion and hardness as level increases, homogenization of individual instrumental outlines, coarseness and truncation of reverberation decay, lack of ease, synthetic character to tonal colors, lack of textural differentiation between instruments, and a general uninvolvement in the music.

No, the HDCD conversion wasn't completely transparent—there was a slight loss of space and inner detail—but the difference was remarkably small considering that the signal was being digitized and converted back to analog. More fundamentally, however, the music had a wonderful smoothness, ease, grace, resolution, involvement, and effortlessness radically uncharacteristic of digital. I've heard preamplifier line stages that introduced more coloration than HDCD processing. Had I not heard HDCD myself, I would have regarded as preposterous the assertion that an entire digital encoder/decoder could be more transparent than some high-end line-stage preamps.

The next comparison was of HDCD encoding without the optional HDCD decoder. Encoded discs can be played back without the decoder and still realize some of HDCD's sonic benefits. This was a fundamental prerequisite of HDCD; without backward compatibility with the huge base of existing players, HDCD would be confined to a small market of dedicated enthusiasts. The difference between standard decoding and HDCD decoding was significantly larger than the difference between the analog master tape and the HDCD encode/decode chain. Throwing the switch to standard decoding shrank the recorded acoustic, brought the presentation slightly forward, blurred the timbral and spatial distinctions between individual instruments, and sounded more "digital." Nevertheless, many of the qualities heard with the full encode/decode cycle were apparent, and the non-HDCD decoded signal was still appreciably better than that produced by conventional A/D converters.

Finally, we heard what HDCD could do when the encoder was driven by a live microphone feed. Keith played an HDCD-encoded DAT recording of the CD Testament (Reference Recordings RR-49CD) with full HDCD decoding. The sound was stunning, with depth, space, the ability to distinguish individual instruments from the whole, beautifully portrayed timbres, and an overall sense of ease and involvement I've never heard from digital. JA said to me later, "I can't believe we were listening to a DAT!" Although this HDCD-encoded disc is now available—and sounds terrific without HDCD decoding—it will be some time before any HDCD decoders find their way into high-end digital processors.

The most startling realization was that with full HDCD encoding and decoding, it will be possible to have a source for home music playback that is very nearly the equal of Keith Johnson's first-generation analog master tapes played back on his own tape machine. No, it wasn't identical, but the digitized version was stunningly close to the analog source, and orders of magnitude better than any other digital system I've heard—Sony's Super Bit Mapping included (footnote 1). With the HDCD encoder driven by the live mike feed, however, the digital version may even be superior in some respects to the analog tape. The digital master won't suffer from modulation noise, speed fluctuations, and other problems inherent in analog tape. I'm sure the HDCD process has its own artifacts, but it remains to be seen which format's anomalies are more sonically benign. This question, however, is academic to consumers who don't have access to first-generation analog master tapes; CDs made with HDCD will certainly sound better than LPs.

How does HDCD work? That's being kept a secret. Whenever the designers are asked any technical questions, they respond with very carefully guarded—no, evasive—answers. Many key patents are still pending, and the inventors want to reap the rewards of their six-year research and development effort. The system now takes up a table top and is a serious piece of engineering. I speculate that the A/D converter is an oversampling type and that the processing takes place at a high sampling rate and long word length, which are then further processed to be compatible with existing standards. The process may also use techniques such as noise shaping and subtractive dither. Part of HDCD's musical performance is no doubt due to the extraordinary care taken in implementing conventional technology such as the analog stages, A/D converter, D/A converter, and just using good overall engineering. For example, the analog stage uses no printed circuit board; instead, the circuit is built in three dimensions, with the most meticulous and beautiful soldering I've ever seen. The system we heard is likely the most fully realized digital system extant.

With HDCD, digital audio has passed an important threshold: It can now hold its own with—or even surpass—analog for musical performance. What's really exciting, however, is that this level of sound quality is possible within the confines of existing format limitations—the 16-bit, 44.1kHz bottleneck foisted upon us by CD's inventors, which was all that was commercially realizable back in the late '70s (footnote 2). This compatibility factor could make HDCD a reality for the masses, not just an esoteric system which can be heard only with nonstandard hardware (footnote 4).

In addition to the musical benefits, the existence of ultra-high resolution source material will energize the entire high-end audio industry. There will be much greater incentive for designers to improve their products and for listeners to upgrade their playback systems if the source contains vastly more information than had been available (footnote 3).

Previous reports of better A/D converters and Sony's Super Bit Mapping have been incremental improvements; HDCD is a wholesale leap forward, in my opinion. If HDCD is widely adopted by mastering studios (the reaction thus far has been very positive) and the decoders are readily available in digital processors, the benefit to the music lover and audiophile is incalculable. For the first time, consumers can play back in their homes a source that is nearly the equal of first-generation analog master tapes played on a state-of-the art tape machine. In some respects, the purely digital source made by feeding the live mike signal into the HDCD encoder may be better than the analog master.

The time has come for us to view the "Perfect Sound Forever" pronouncements of ten years ago as equal in absurdity to Anna Case's assessment of Mr. Edison's machine.

---

Footnote 1: See "Industry Update" in this issue for my follow-up on Sony's Super Bit Mapping technique.—Robert Harley

Footnote 2: When you consider that it took Sony's commercial might and the fact that Sony had the patents on the essential error-correction scheme to persuade Philips to change from their original 14-bit CD standard, you rightly shudder at what might have been.—John Atkinson

Footnote 3: One drawback to all such improved-definition CD systems is that they need good DACs to realize the improvement in sound quality. As I said last month, play back an HDCD-encoded/decoded CD, or even a Sony SBM CD, on a cheap, poorly aligned CD player, and you'll still get unmusical, cheap CD sound.—John Atkinson

Footnote 4: Pacific Microsonics was purchased by Microsoft in September 2000.—John Atkinson