By John Atkinson • Posted: Jun 9, 2007 • Published: Mar 9, 1990
Stuck out here in the desert depths of the Southwest, we look forward to visits from out-of-towners. So when David Wilson, one-time audio reviewer but now full-time high-end manufacturer, called to say he was going to be in Santa Fe, there was a flurry of activity. David had agreed to an interview, so I started going through back issues of The Absolute Sound and Stereophile for background. Vol.6 No.2 of Stereophile from 1983, with its front-cover photograph of David and Sheryl Lee Wilson with their WAMM speaker system, seemed a good place to start—except that nothing inside the magazine corresponded to the cover picture. It was the next issue that had featured Larry Archibald's write-up on the WAMM, and once I opened its pages, I got trapped into reading the entire issue.
J. Gordon Holt's "As We See It" involved an intriguing LaserVision disc. An orchestral recording, it had been balanced in a manner that, had it been a sound-only disc, would have destined it for the circular filing bin. As each instrument took a solo, the sound engineer had boosted its level so that it dominated the mix. However, the video disc revealed the aural spotlighting to coincide with close-up shots of the respective performers, and the venerable JGH (in whose ears we trust) felt that, owing to the dominance of sight over the other senses, such a distortion of the soundstage was subjectively acceptable. It was only in the absence of the visual sense, he argued, that listeners become critical of flaws in soundstage reproduction, the solidity of the sonic-only experience being destroyed by a heavy hand at the mixing console.
Which brings me to a number of letters this month concerning the reproduction of the recorded soundstage. A Mr. L. A. Whitcher accuses me of muddled terminology when I recently referred to a "possibly three-dimensional space" being created between and behind the loudspeakers. "If it's not three- dimensional, it's mid-fi, no matter how much you paid for it," guffaws Mr. Whitcher, obviously thinking that I hardly know which way is up.
What we have here is dimensional confusion. Many people think that the three dimensions that define a reproduced soundstage are "Left," "Right," and "Back," in which case any system that threw image depth would be "three-dimensional." This is incorrect, however. The three classical dimensions, each defining motion at a 90° angle to the other two, are "Length" (left–right), "Width" or "Depth" (front–back), and "Height" (up–down). A system that reproduced a stereo image with impressive depth would therefore be reproducing a two-dimensional stage. If its image had no depth, then it would be one-dimensional. (A one-dimensional object is a line, having length but no depth/width or height.) This is the context in which I had referred to a "possibly three-dimensional" soundstage last November, not one which could possibly have depth—which I understand to be Mr. Whitcher's interpretation—but one which possibly gave a faithful sense of image height. Although many audiophiles would insist that their systems do reproduce a sense of height, when I have experienced image height, it has nearly always turned out to be spurious, due to system flaws. The fundamental problem is, you see, that nearly all the microphone techniques recording engineers use to record music are incapable of capturing any height information.
The argument is that height perception is the result of system flaws because current microphonbe techniqueshniques
A common philosophical trap fallen into by audiophiles is to assume that any LP or CD inherently contains within it the information necessary to recreate the live illusion. This just isn't correct. Even the finest system will not create a soundstage with an accurate sense of height unless care was taken to ensure that the appropriate height information was captured at the recording session. I am not saying that it is impossible for a stereo system to throw a sense of height—the new Chesky Test CD that I mention in my loudspeaker reviews this month contains tracks where the signal can reproduce as being above the plane of the loudspeakers—but that conventional stereo microphone techniques do not capture the aural clues that allow the ultimate listener to perceive height.
There is one possible exception to this blanket dismissal: the tiny fraction of recordings where the engineer has used a single-point stereo technique. As implied in the addendum to Professor Peter Fellgett's letter, if early reflections of the direct sounds of instruments are captured by a stereo mike technique without any lateral spatial distortion—ie, all the components of the reverberant field come from the correct lateral positions between the loudspeakers—this "ambience-labeling" will contribute to a feeling of solidity and depth to the reproduced soundstage. And as the reflections of the direct sounds from the floor and ceiling of the recording venue will also come from their correct lateral positions, it can and has been argued that this will lead to an accurate reproduction of image height.
This general philosophical confusion also applies to the width of the soundstage, a topic examined in three letters from Singapore reader Yip Mang Meng. Some writers have naÏvely expounded that if the live orchestral image extends from wall to wall, then so should that experienced from a hi-fi system. As with image height, however, the ability to throw a soundstage that extends beyond the speaker positions is not primarily a system characteristic but one that is fundamentally dependent on the information encoded in the recording. It is misleading, therefore, to imply that a perfect system will inherently reproduce images beyond the speaker edges, as nearly all stereo microphone techniques again completely fail to encode the information necessary to reproduce such a soundstage.
A single microphone may pick up sounds from many directions, but when a single-channel recording of its output is played back, all the soundsources will appear to come from the speaker position. Soundsources that were further away from the mike will have a larger proportion of reverberation captured, thus their images will appear to be further away—it has been known since the dawn of recording that monaural recordings can still reproduce the depth dimension. Lateral imaging, by definition, is a function of the relationships between the signals in two or more reproduction channels, and conventional multi- and widely-spaced–mike techniques produce images that must by definition fall at or between the speaker positions.
Consider a spaced-mike technique, that used for Mercury "Living Presence" and Telarc recordings. All sources to the sides of the mike positions will be captured with a somewhat different ratio of amplitudes between the two recording channels, but with a constant time delay. When that recording is played back over two loudspeakers, those side images can't help but be localized in the speaker positions owing to that time delay. The soundstage thus remains bounded by the speakers.
However, while I was thinking about the abilities of the various mike techniques to reproduce height information, it struck me that perhaps the classic Blumlein pair of figure-8 microphones, vertically coincident and angled at 90°, should capture the information necessary to reproduce image positions beyond the speakers. Soundsources in the angle subtended by the mikes will produce in-phase signals in the two microphone outputs that differ in amplitude: this mechanism defines all the traditional image positions between the loudspeakers. But consider a source to the left-hand side of the microphone array outside this angle. It produces identical amplitudes in the two mike outputs, but whereas that in the left-pointing mike will have positive polarity, that in the right mike will have the opposite polarity. When the two-channel recording is played back over loudspeakers, therefore, such a soundsource will produce out-of-phase information in the speakers with the phase-lead to the left loudspeaker. This is exactly the interchannel information relationship required to produce an image beyond the left-hand loudspeaker position.
To test this hypothesis, when we recorded the music for the second Stereophile album—see "The Final Word," p.226—I also used a Blumlein pair of figure-8 mikes to record Larry Archibald mapping out the soundstage from way beyond the mike position on the left side of the church all the way to the right-hand wall of the church. (The "traditional" soundstage covered by the mikes occupied the center third of the church.) When the tape was played back over a pair of Thiel CS5 loudspeakers—these superb speakers create one of the best-defined soundstages I have heard—Larry's voice and handclaps could be heard way beyond the speaker positions, just as predicted.
Now as far as I know, no one records instruments and voices outside the "traditional" pick-up area of a Blumlein pair. Their images should therefore stay within the bounds of the speaker positions. But the ambience and reverberation recorded in such a manner should extend beyond the speakers, adding to the sense of realism and image.