Tam Henderson, Reference Recordings Page 4

The reason CD has become the dominant medium must have to do with the lack of noise, the lack of ticks and pops and groove skips, etc., that we are too well acquainted with. And the relative permanence of CD is certainly very attractive. It's convenient, it sounds the same for many years, it doesn't deteriorate.

Atkinson: And you don't notice the lack of what you should be getting.

Henderson: Precisely. You get accustomed to the sound; that sound is satisfying by and large, so it's good enough.

Atkinson: If you say that some of the problems with digital sound are due to the editing machine and perhaps some are due to the digital recorder, then you have just concluded a project where you bypassed most of these problems by recording direct-to-CD!

Henderson: The way this whole thing started was when a pianist friend of mine asked me a couple of years ago to come and see this computerized reproducing piano that Bösendorfer has. I wasn't enthusiastic. "Oh, a player piano. So what?" He said "No, come and listen." So we did. It was amazingly good; it sounded nothing like the mechanical reproducing pianos that I've heard. This one was darned convincing. I'd be hard pressed to say that there was not a real body there playing.

Atkinson: It captures all the subtleties of touch...

Henderson: Nuance. Yes. And phrasing, yes. It's amazingly successful. We said "Okay. It's successful. So what? It's still a player piano. Why do we need a player piano?" So Keith and I began to scratch on this—but let me jump to what is necessary to cut a CD.

After you send the tape to the mastering plant, the tape is analyzed as to where all the cue points are. You have to know down to one thirtieth of a second where the tracks begin, where they end, and so on. You have to enter all of these points into a computer and it encodes them so that you have a directory at the beginning of the CD.

Well, this Bösendorfer piano has all sorts of editing resources: you can correct wrong notes, you can speed it up, slow it down, even transpose. Once you have the performance in its computer, there's a lot of flexibility in what you can do. So, say we have our edited program put together, lasting x number of minutes. The instrument will repeat itself exactly every time. So if we made a digital tape as we normally would of the performance, complete with time code, and send that to the plant, they can analyze where everything is. And then we could send them down the program for real and cut the disc from it.

Conceptually, that's the way it would work. It isn't complicated. But implementing it turned out to be rather complicated. I won't go into all the problems that we hit in trying to do it. But essentially we had a microwave link at our end, and a microwave link at the CD manufacturing plant where the disc was cut. The signal had to bounced off Mount Wilson in Los Angeles because there were buildings in the way. The player piano was at our facility, playing itself, in the hall, with the mikes feeding the digital processor. The digitized signal was then sent to the mastering plant and they cut the [glass master] as the piano was playing. They already had worked out all the cue points ahead of time.

The great sonic benefit of doing it this way is probably not avoiding the tape itself. We don't know, but we think that probably is not as much the limiting factor so much as the editing. We bypassed all of that equipment; I think this will be as clean a sound as digital is now capable of generating. There's nothing in the chain except the digital convertor at our end. That's the way it was done.

At the same time, we cut the world's first direct-to-disc DMM LP—which was also more complicated than we expected. The reason a direct-to-disc DMM cannot be made is that the lathe's onboard computer has to see a preview of 0.9 second. With a conventional, lacquer-based LP cutting system, the groove spacing and depth can be adjusted manually. But with the DMM system that's not available; you must have this preview so the computer can analyze the signal and tell the cutting stylus how to behave. So with this reproducing piano we were, with a lot of whiffing around, able to get a 0.9s preview for the computer to cut the groove.

We had to revise all of our thinking about this because the plan we had when we arrived at the session didn't work, and we had to invent another way to do it. Without the three channels of Keith's machine, it couldn't have been done. The piano, or the computer that drives the piano, has a digital datastream that turns out to be a single channel of information that can be recorded either digitally or on an analog system. That is accessible. We recorded the datastream of the edited program from the computer, one channel, on to one channel of a digital tape. Okay, that signal is then also fed to Keith Johnson's analog machine which acted as a time delay.

It turns out that the time delay between the record head and the playback head of his analog machine is about 0.3 second. So [the Bösendorfer control signal] was recorded on one channel, taken off the output of that track, put back into the second track, taken off the output of the second track, and put on the third track. By the time we got to the output of the third track, we had our 0.9 second delay, and its output then went onto the second track of the two-channel digital tape.

Now we had the piano-encoding information on two channels, one set 0.9 second behind the other. We then played back the first of those tracks to the player piano, miked its performance, and recorded those two channels onto two channels of Keith's analog machine, with the 0.9s-delayed computer control track rerecorded on the third track of the analog machine. The three-channel analog tape is our master: the two channels of audio information are fed to the cutting lathe to provide the two channels of preview it needs to see. 0.9 second later, the piano plays the same information, this time the mikes feeding the lathe with the program.

There we were, cutting a direct-to-disc LP session. The computer was playing the piano, the preview provided the disc-cutting computer with its information, the disc was being cut, and we all sat around reading the Sunday paper! I had to get up about every three minutes and make a spread on the LP. And we have 30-plus minutes of music on each side of our direct-to-disc LP.

The music, incidentally, is Dick Hyman playing music of Fats Waller—stride piano music of the '30s and '40s. It's the same performances on both direct-to-CD and direct-to-DMM. I had insisted that this high-tech gimmick recording also be musically worthwhile, so we went to Dick Hyman (who is well known as one of the great stylists of the music of that era) and asked him what he would like to do.