Interviews

I think this record will be twiddling its thumbs five years from now, waiting for equipment manufacturers to make real systems for real people—as opposed to special custom-made commercial systems—that can actually play it. It's not hard to track, but the tough part is to reproduce the full power in the bass and midbass. There are very few systems around that can actually do that. The dynamic peaks are tremendous, +19dB over zero recording level. The cutter can legitimately put more energy in the groove, but Sax is a world-renowned expert in disc-cutting and he knows what real-world cartridges can really do. He sets his levels so that the highest peaks will be trackable by anybody. But then the average level goes down. The low average level shows that it's a wide dynamic range recording.

My complaint about current digital could be summed in two words: "Premature Standardization," except that "premature" implies that at some time you will be able to do it right, and that is a big question mark. I must make it make clear that I don't really feel like a voice crying in the wilderness on this subject. The BBC recently came out with a report intended to answer the question "How many bits do you really need to handle music as it comes straight from the microphone without any compression, limiting, or dynamic processing?" They came up with 22 bits per sample—the current standard is 16. This is inconceivably higher resolution than anything we have available now, or anything that companies are even dreaming of. Quite apart from the fact that we already have a standard.

Atkinson: It is said that 16 bits is enough if you know everything about the signal: where it's going to be loudest, and how loud; where it's going to be softest, and how soft; and, most importantly, if that is the final medium, for playback in the home. As I understand it, if you're in a recording situation and you need to mix and play with the signal at all and still want to end up with good 16-bit words, you need to start with, if not 22, then at least 18.

Boyk: There is something I would like to make clear. When you add a bit to 16, you're not adding a difficulty of one sixteenth, you are doubling the difficulty; so when you add six bits, you're making it two to the sixth—64—times as difficult. But it is not 64 times as difficult, it's a zillion times as difficult because there isn't anyone in the world who is even dreaming of 22-bit conversion at the kind of rate required for audio.

Let me indicate where the state-of-the-art stands. I called up a well-known manufacturer of converters for CD players, and asked what their fastest 16-bit converter was. They make one that will operate at 110kHz or 120kHz, so I asked if it was really linear at high precision. I'm glad I asked this question; the answer was that, if I wanted a high-precision 16-bit converter, the fastest rate it would operate would be just 6kHz! Which means you're looking at an effective signal bandwidth of just 1.5kHz, if you really want to handle things right. "What about these ones that go at 44k, for CD players?" I asked. "They're not as accurate," came the reply. You ask when they will have high-precision 22-bit converters, and they start laughing.

However, I don't believe that 16 bits is adequate, even if you do know everything about the signal. I think the BBC said that it's adequate if you're going to broadcast 13 bits, which is what the BBC does. Then there's enough extra places to keep the accuracy of the 13 bits.

Now what if someone said that they had a machine that does 22 bits per sample at a 200kHz or 300kHz sample rate, with a truly linear—not just monotonic, which is an incredible cop-out—performance? I'm perfectly willing to believe that, if implemented correctly and surrounded by analog electronics of the highest level—and when you look at the requirements for that, it's incredible how good they have to be—it would be excellent. I don't expect that I will hear that in my lifetime. But suppose that it were excellent, suppose, magically, we had it here now—now a Compact Disc holds just eight minutes of music! Gee, eight minutes of music and an LP is as good and it can hold 40 or 50 minutes of music! Hey, an LP is a pretty dense storage medium, isn't it?

The answer is that it is a dense storage medium; it's a fantastic storage medium—it's our best! The proof of that is Sheffield's recordings over the last few years, where you have the LP and nothing else. Anybody who wants to hear what woodwinds sound like, go listen to the Sheffield Chicago Symphonic Winds record, the Mozart and Grieg. That is the sound of musical instruments. It's just extraordinarily transparent to the music. I would even make that claim, to almost the same degree, for a couple of my records on Performance Recordings. There just isn't anything to get in the listener's way with the Prokofiev Sonata record or the two we've made since then. If there were, I would have got rid of it (although I would like the hiss level to be lower).

Atkinson: Can we talk about the slogan on your Performance Recordings T-shirt: "Digital finishes what the transistor began."?

Boyk: It comes back to the fact that some of the problems of digital are masked in typical studio practice by what's wrong with typical studio equipment. Condenser microphone diaphragm resonances can mask the additional harshness that digital can produce; the mikes don't have particularly good ambient pick-up, owing to their solid-state electronics, which masks digital's loss of ambience; and the mixing desks are mostly pretty bad.

There's been a lot of play recently about how wonderful some of the commercial recordings from the early stereo era of the '50s are. I think that's right; a lot of them were wonderful records, and it's significant that that was an all-tube analog era. People used high-resolution equipment then. In the '60s, however, mastering engineers and studio engineers started getting early solid-state power amplifiers to drive their loudspeakers, and the resolution went out the window. That has nothing to do with what solid-state is ultimately capable of, or what we have today, but solid-state amplifiers, we would all agree, were then a complete disaster!

I'm not just saying merely that I dislike what solid-state tends to do, and what digital definitely does; what I'm saying is that the one really leads to the other, that the degradation of solid-state is a step on the way to the degradation of digital. I believe that if you went in one jump from the state-of-the-art from 1958—stereo, but tubes and analog—to today's art—digital and solid-state—that leap would be absolutely unacceptable. It's only because it was first mediated by the transfer to solid-state that people accepted it.

I gave a talk at Caltech to explore that idea. It was called "1960 vs 1985 in Recording" and, just so nobody was in any doubt as to where my own opinion lay, it was subtitled "A Quarter-Century of Degradation." I went to a used-record store and bought actual LPs that were 25 years old. God knows what their history had been. I played them on a decent, but not great, turntable/arm/cartridge combination; my idea was to play the same piece of music from the same record label in a current all-digital CD, played back on a Meridian MCD player through the same fine system, to compare what these companies thought was the way to do it then with what they think is the way to do it now.

I played three such comparisons—the LPs came from 1958, 1962, and 1965—then asked people if, assuming these three comparisons to be the Universe of such comparisons, did they feel that there was a difference between how well the music was served by the two technologies? Everybody said that there was a definite difference, no question about it. I then asked who felt that the technology of today—solid-state, digital—serves it better, and five hands went up. "Fine," I said, "How many people feel that the art of a quarter-century ago served the music better?" 46 hands went up! Well, that's a stunning indictment! (footnote 2)

Atkinson: You seem to have ended up with conclusions which run counter to accepted wisdom. You make records with 20-year-old ribbon mikes; you think amplifiers sound more neutral when using tubes, and let the music through more benignly; you've proved, at least to your own satisfaction, that the highly touted digital recording system is not working as it should. How do you convince people that you're not saying all this from some reactionary stance?

Boyk: Oh, the convincing is easy! Just let them listen to one of my All-Tube Analog records on a well-chosen playback system—by which I don't necessarily mean an expensive one. The Sheffield Firebird is all-tube; their Lincoln Mayorga/Arnold Steinhardt Dvorak and Strauss violin and piano record is all-tube. When you say that I run contrary to the received wisdom, in some sense that is true, but in another it's not: if you look at the record companies that are internationally acknowledged as making the sonically finest records—Sheffield, Telarc, Reference Recordings, Audiofon, Wilson Audio, Harmonia Mundi, Hyperion, Meridian in England, Proprius and Opus Three in Sweden, (I'll modestly include my own company, Performance Recordings, but we're microscopic even when compared with these companies)—only Hyperion and Telarc are committed to digital recording. Over 80% of the highest-quality companies believe in analog for recording music.

So where is the received wisdom?

Atkinson: But when you look at these companies who are respected for their sound quality, their total sales are but a minute percentage of the sales of a company like Deutsche Grammophon, Philips, or CBS.

Boyk: We're not talking economics here, we're talking fidelity.

Atkinson: But the received wisdom is laid down by the people with loudest the voices.

Boyk: Like Galileo, I say it still moves!

There's no question, however, that commercially it is all over. Commercially, it's all digital. Analog is dead.

For now!

---

Footnote 2: The introduction of digital is like nothing so much as the introduction of transistors 25 years ago. If you looked at the situation in 1970, or even 1965, it was all over for the tube, it was all transistors. But a couple of years ago The Wall Street Journal had an article about the renaissance of tubes! In the last decade 20 or so companies have come into existence to manufacture tube equipment. Maybe the same thing will happen with analog; or maybe digital will get good enough.