Sheffield Steel? Doug Sax

Doug Sax is undoubtedly one of the most controversial and outspoken figures in audio. As co-founder, with Lincoln Mayorga, of Sheffield Lab, Doug pioneered the first modern direct-to-disc recording. His perfectionist methods may be controversial, but the results certainly are not: Sheffield Lab recordings are nearly universally praised by the audiophile community, while the Billboard Hot 100 always features at least one Sax-cut disc.

Doug was an early champion of tube electronics and a vociferous critic of digital audio. In fact, he once distributed T-shirts with the inscription "Stop Digital Madness." While at Disctronics, I worked on the CD mastering side of many Sheffield projects, and consequently had several interesting digital audio discussions with Doug. To share Doug's fascinating ideas and opinions with Stereophile readers, I met him at The Mastering Lab, his disc-mastering facility in Hollywood, and asked him what, specifically, did he find wrong with the sound of CD?

Doug Sax: I think that the Compact Disc is still in its infancy. I don't think you are getting a 16-bit product. You are getting maybe 14 bits out the door. Professional recorders are marginally 16-bit, but off the tape, in video storage, I doubt it (footnote 1). Have you ever been in a professional, high-quality TV station? It's sort of a shock when you look at a live monitor. What you have there is 525 lines. You've got the real NTSC, and when you get home to your TV, you say "This is chopped liver!"

I think that compares with what can be done in a 16-bit, 44.1kHz linear system. Current digital recordings are probably on the level of a home TV compared to that studio monitor. There is much to be done to bring it up to what it's capable of. It's strange because the home format [CD] has virtues over the professional format [CD master tape]. It's a total reversal of where we'd been in audio up to the introduction of CD. The professional master tape was always stunning compared to the home format which is a generation down, a mass-produced pressing that wears out as you look at it, is hard to play back, with compromises made to get it [the recorded sound] into your home. The CD has no wear: optically verified that it has all the dots and dashes, when it's optically verified two years later it still gives the same read-out. The professional format in digital form is a video-based tape that gives you more errors the more you play it. It may be good for 25 plays, but where do you draw the line? No question that you're going to play it 50 times. No way. You cannot make a data-to-data copy that is inaudible. I'll stake my reputation on it.

Home storage is now dramatically superior to the professional format. So it's an opportunity. Everyone seems to think that oversampling playback is better, is the way to go. But almost no one, with a few exceptions, has an oversampling CD on the record side in their possession. You also have the fact that you're "polishing the tail." You're at home and you have your fancy CD player and its 8x-oversampling or a million times oversampling and its la de da de da—but it's not been done to the recording. That's gone through stonewall 22kHz filters, through an A/D converter that will not measure and give the true numbers of a 16-bit converter.

Robert Harley: How then do you account for CD's enormous popularity?

Sax: Their strongest point is that they play back at the same level with the same frequency response. Probably the easiest thing in the world to do is to play an LP record incorrectly.

Harley: With CD you don't have the vagaries of cartridge alignment.

Sax: A thousand vagaries. That makes evaluating what a record can do secondary to being smart enough to put together a system properly aligned, components happy with each other, that could play what's on a phonograph record. The easiest thing is to play a record wrong. I go around to high-end stores and hear records played abominably; and they're the people selling the equipment. You always read about the average guy on the street who says "I got my CD player and can't believe how great it sounds. What an improvement over records." Now, you've got some guy at a high-end magazine who says, "They still don't have the music I have on my best records." They are both true statements.

Harley: The guy on the street has a hundred-dollar turntable...

Sax: ...and never got what's on the record. It's very difficult to get what's on the record. And for $195 you can buy a player that will get 80–90% of what's on that CD. Many of these even sound better than some of the $2000 audiophile players. Now I can say that because I've heard some $2000 audiophile players—they have oversampling, all the buzz words—that didn't do what was on the tape. I made the tape and I would expect that, at the minimum, their playback would be as good as my professional playback.

Harley: How many of the problems of CD can be attributed to the input section of a Sony PCM-1630 (footnote 2)? It uses half a dozen op-amps in the signal path, including 5532s.

Sax: 5532s, properly implemented, are better than most of what you've got in your hi-fi, believe it or not. They're not the fault. But the complexity of the circuitry is to be questioned enormously, as is the fact that the 1630's not oversampling, that it's not a very linear 16-bit converter, that it has between two and ten times the distortion of your home playback system. If you look at your CD player and say "Well, look at this, 0.01% distortion, 95dB S/N." But that's off a code from a computer put on that disc. It's not an in-through-out measurement of the system. Because you don't get these numbers from an A/D that's in a 1630 or a 1610 or a Mitsubishi or any other professional recorder. You don't get these numbers. You don't get your 0.002% distortion, because the A/D converters won't do it. You don't get your noise numbers that would even show a true 16-bit system. And after a couple of generations of video tape, those numbers will get worse because you're not getting all the data over. Martin Colloms will tell you that the biggest factor is that the time in the video machine is skewed. The numbers are there but they're not in the same time [relationship] they were recorded and it doesn't sound the same (footnote 3).

Harley: Do you mean jitter?

Sax: Yes. You know that this signal is converted to FM, to get put on to video tape. Then it gets decoded, then it gets clocked out, and the deck itself is full of jitter. It's very primitive storage. But look how good the home storage is; it's not going through FM. The decks themselves would make this stuff better than it is. The players, however, vary enormously in their jitter, and even though they'll measure the same distortion off a disc, they won't sound the same on music.

Harley: So you think the weak links are the A/D converters and the tape transports?

Sax: Those are probably the grossest links. And the filtering and sample and hold circuits. And this process changes the sound. It becomes "digital." I think that if you had a well-conceived 2x minimum oversampling, truly linear, 16-bit system with no degradation in the storage, you wouldn't recognize it.

Harley: Do you expect digital to get better?

Sax: I expect it to get better and I'm working to make it better. One of the things you have to do to make it better is you have to know that it's not right. All you have to do is say "this is terrific" and then you don't improve it. In other words, realize it's here to stay but be critical of it. We're possibly going to construct and offer for sale an A/D and D/A box that interfaces AES/EBU (footnote 4) so you can take your 1610 or your DASH or whatever, use it to store, but we'll handle the signal path from A to Z.

Footnote 1: The worldwide standard for CD master tapes is a ¾" U-Matic video tape. In this format, digital audio is stored as a video signal: white represents binary one and black binary zero. Each horizontal line of video stores 193 bits.

Footnote 2: The music on the majority of CDs, especially ones made from analog sources, passes through the analog input of a Sony PCM-1630. The 1630 low-pass filters, digitizes, and converts the signal to a pseudo–video signal for storage on ¾" tape for CD mastering. See Martin Colloms's review of a 1630 in Vol.10 No.5 for more detail on its design and how it measures and sounds.

Footnote 3: See "Industry Update," September 1989.

Footnote 4: AES/EBU (Audio Engineering Society/European Broadcasting Union) is a professional digital interface format. It is nearly identical to the S/PDIF (Sony/Philips Digital Interface) found on many CD players and DAT machines.