Stereophile Test CD 3 Tutorial Tracks: Microphone Techniques
 Microphone Techniques & Soundstage Maps (DDD) 6:08
2 omnis (Index 1), 3 omnis (Index 2), ORTF cardioids (Index 3), "shuffled" ORTF cardioids (Index 4), Schoeps SFM 6 "sphere" (Index 5)
Producer: John Atkinson
Recording Engineer: Robert Harley
Recording Venue: The First United Methodist Church, Albuquerque, NM
Recording Date: November 5, 1993
Microphones: two Brüel & Kjaer 4006 *" omnidirectionals with black (diffuse-field) grids; as above with EAR The Mic set to omnidirectional pattern; two AKG D190E dynamic cardioids, Schoeps SFM 6 "Sphere"
Microphone Preamplifier: Sonosax solid-state
Recorder: Nagra-D with integral 20-bit A/D Converters
Digital Mixing & Shuffling: Sonic Solutions Sonic System hard-disk editing system
I included this track for two reasons: One, it's a useful primer on the way in which stereo microphone techniques affect the accuracy of what's recorded (footnote 3); two, it will enable audiophiles to check out their systems' soundstaging abilities. During a break in the sessions for Stereophile's next Robert Silverman recording, Robert Harley used different stereo sets of microphones to record me talking and banging a cowbell with a drumstick as I walked first from the far left of the church's 45'-wide stage to the far right, then from the very back of the 65'-deep auditorium up to and past the microphones and piano (fig.1).
Fig.1 The First United Methodist Church, Albuquerque, New Mexico, plan view. Arrowed line shows paths taken by JA while recording Test CD 3's soundstage-map track.
What you should hear: The first stereo miking array consisted of two Brüel & Kjaer 4006 *" omnidirectionals, spaced about 10' apart and pointing toward the roof. Used with the black (diffuse-field) grids, this results in an overall flat response. As well as hearing the cowbell, my voice, and my footfalls illuminating the church acoustic, you should perceive me mapping out a U-shaped soundstage. When I start at far stage left, my voice isn't that different in level in the two channels, because the distance between the two microphones is smaller than my distance from the closest one. However, because the sound of my voice reaches the nearest microphone about 7 milliseconds (ms) before it does the other one, this time delay locates the image in the left speaker position. As I'm a long way from the microphones—about 16'—there's a bloom of reverberation surrounding my voice that moves the image significantly behind the speaker. The image is also quite wide.
Schoeps SFM 6 one-point stereo microphone
As I walk toward the microphone, the interchannel time delay, though decreasing, keeps my voice at stage left. The decreasing amount of reverberation, however, causes the image of my voice both to get smaller and to move closer to the listener until I reach the left-hand microphone, when it should be positioned right at the speaker position. Then, as I walk between the microphones, the rapidly changing time relationship between the two channels causes the image to move very rapidly from the left speaker position to the right.
Once I reach the right-hand microphone position, the process repeats in reverse, giving the familiar U-shaped soundstage produced by a pair of spaced omnis. Depending on the microphone spacing and hall acoustic, the "arms" of the U can bend back toward the center; but in this case, I arranged things to give a true U shape. If you don't hear precisely what I've just described when you listen to this track, then something in your system or room is impairing the imaging accuracy.
The second half of this demonstration enables you to check out your system's ability to throw accurate image depth. As I run to the back of the church's nave, you should hear my footsteps, roughly placed in the center of the stage, rapidly receding. As I walk back to the piano, you should hear the image of my voice and of the cowbell moving from somewhere behind the wall of your listening room up the plane of the speakers when I announce that I'm level with the microphone positions. Because of the center-stage instability I mentioned earlier, my small movement of walking around the piano past the keyboard should be heard as a lurch in the image position all the way to the left speaker position, before I then recede a short distance behind the loudspeaker plane. (No stereo microphone technique can distinguish soundsource positions in front of the mike array from those behind—all image depth is therefore perceived as being behind the speakers.)
The pulling of the central images to the sides with a pair of spaced omnis is why the classic spaced-omni mike setup—as practiced by Mercury, Telarc, and Everest—includes a third, centrally placed mike to stabilize the soundstage center. To demonstrate the improvement this brings, I mixed in the output of a third omnidirectional microphone (this time an EAR tubed model) positioned halfway between the other two. Compared with the two-mike recording, the soundstage is still U-shaped, but now the image positions between the speakers should be more stable, and the image lurch in the depth section of the recording should be less severe.
For the third demonstration, Bob used a pair of microphones with a cardioid or directional pickup pattern in an "ORTF" configuration. The microphone stand was placed at the center of the stage, where the third omnidirectional mike had been in the previous example. "ORTF" refers to a technique devised by the French broadcasting organization (Office de Radiodiffusion-Télévision Française) whereby two cardioid microphones are angled at 110 degrees and spaced apart about 7"—the average distance between a human's ears.
The two directional microphones basically encode the directions of the voices and instruments by the different loudnesses they pick up. In itself this would give a narrow stereo image—"fat mono" is how one writer described it—but by spacing the microphones apart, a little time information is added which ensures that the image extends evenly across the full spread of the loudspeakers. (Sound reaching the microphones from the left, for example, will reach the left-facing microphone approximately 0.7ms before it reaches the right.)
You should hear my image move smoothly and evenly from beyond the left loudspeaker position, across the center of the stage, to beyond the right speaker position. Because the microphones do not pick up sounds to their rears, there should be noticeably less hall ambience audible than with the omnidirectional-based techniques. As a result, the soundstage will be somewhat foreshortened. Cardioids tend also to have rolled-off lows: the traffic noise is less obtrusive with this version than with the others. In addition, because even good cardioid microphones have difficulty maintaining their directionality at low frequencies, there's a tendency for the low-frequency and lower-midrange soundstages to be narrower than that at higher frequencies, leading to a loss of "bloom" with music recordings made in this manner.
The next demonstration shows how this can be adjusted by a bit of judicious post-processing. "Shuffling" was first described in the '30s by the British engineer Alan Blumlein (footnote 4). The first time I was exposed to its use to widen the recorded soundstage was in a lecture in 1987 by British mathematician Michael Gerzon, based on an article in Studio Sound magazine (footnote 5). Basically, you process the two stereo channels to produce a pair of signals—one consisting of the sum of the two originals, the second the difference between them. By boosting the bass content of the difference information, you add low-frequency spaciousness or bloom to the soundstage when the sum and difference signals are recombined (fig.2), but without smearing the image stability or accuracy.
Fig.2 Stereo shuffling to widen the low-frequency soundstage, block diagram.
The final soundstage map was made with a single-point stereo mike: the Schoeps SFM6 "sphere," which consists of a head-sized plastic globe with flush-mounted omnidirectional microphone capsules pointing to the sides. The SFM6 is designed to combine the imaging accuracy of angled coincident directional microphones with the low-frequency bloom and spaciousness provided by spaced omnis. It also gives excellent out-of-the-head imaging—almost binaural in its effect—when auditioned over headphones.
[11 & 12] Robert Whyte: "Christe qui lux es et dies" (ADD) 4:56
Musicians: Halcyon, conducted by John Hoban
Recording Venue: St. Jude's Church, Southsea, England
Recording Date: March 20, 1982
Recording Engineer: John Atkinson
Microphones: "Shuffled" AKG D190E dynamic cardioids in ORTF configuration
Microphone Preamplifier: Prokit 6-2 mixing console
Recorder: ReVox B77 two-track open-reel recorder at 15ips with dbx II noise reduction
Tape: TDK GX
Transfer to Digital: ReVox PR-99 open-reel recorder, AudioQuest Lapis balanced interconnects, Manley/UltraAnalog 20-bit Analog to Digital Converter; mastered and redithered to 16-bit resolution after "shuffling" using the Meridian 618 Mastering Converter
Digital "Shuffling": Sonic Solutions Sonic System hard-disk editing system set to 20-bit resolution
Recorded live at a 1982 concert to benefit the UK's Mary Rose Trust (The Mary Rose was a Tudor warship recovered more or less intact with its contents in the early 1980s from the sea off England's south coast), Halcyon was a group of accomplished amateur English singers gathered together for this one concert. Musicologist Bruno Turner wrote that Robert Whyte (1538-1574) was one of the last English composers to write in "the old florid Latin manner." Nevertheless, this unaccompanied setting of the old Compline hymn—one of several he made—is sublime in its simplicity.
I used relatively modest equipment to make the recording, but I felt it worth including on Test CD 3 to illustrate the benefits to be gained by careful use of post-processing—in this case the use of the Sonic Solutions hard-disk editing system to "shuffle" the recording, as explained earlier. In theory, the sum signal should be processed by the inverse of the amount of bass boost to preserve the stereo signal's low-frequency balance. But, as the microphones used to make this recording have a significant bass rolloff when used a long way away from the soundsources, I omitted this.
What you should hear: The individual singers should be unambiguously positioned between the loudspeakers, with the church ambience occupying a dome of sound behind them and to the sides. As with the Empire Brass track (3), when this recording is played on a Home Theater system, the sound of the hall should wrap to the rear, providing an enhanced degree of listener envelopment. As with the following Silverman track, the audience is very quiet, but there is one extraneous noise. A couple of cars also make their presence known.
Track 11, which consists of the first section of the work, played twice, is to demonstrate the choices I had when deciding on the optimal amount of "shuffling." The first half of this track is the original, unshuffled version. It sounds relatively dry, though the imaging is precise. You can easily hear the walls of the hall, but the recorded balance lacks a sense of envelopment.
The second half of track 11 is the most extreme processing I tried, applying 12dB of second-order low-frequency boost below 600Hz to the difference signal and reducing the level of the sum signal by 3dB. It sounds extremely spacious, but the imaging stability is sacrificed. The reverberation also sounds a little artificial and the low-frequency sounds of cars outside the church start to become phasey. You can also more easily hear the church's heating system.
For track 12, I used a less-extreme difference-signal equalization—6dB, first-order boost below 800Hz—and attenuated the sum signal by 3dB, which to my ears gives the best balance between ambient envelopment and imaging stability.
Footnote 3: I examined this subject in depth in a pair of articles in the May and June 1981 issues of Hi-Fi News & Record Review, as did Bruce Bartlett of Shure in the December 1979 issue of db magazine. An excellent book examining the pros and cons of different microphone techniques is The New Stereo Soundbook by Ron Streicher and F. Alton Everest (TAB Books, 1992).—John Atkinson
Footnote 4: See "Applications of Blumlein Shuffling to Stereo Microphone Techniques," M.A. Gerzon, Journal of the Audio Engineering Society, Vol.42 No.6, June 1994, p.435, which has a full list of references.—John Atkinson
Footnote 5: "Stereo Shuffling. New Approach—Old Technique," Studio Sound, July 1986, pp.122-130.—John Atkinson