Another reason that stereo recordings usually sound more natural than mono recordings or stereo mixes made from multitracked mono sources is that stereo recording is often done ambiently-that is, with the mics placed at a distance from the sound source rather than directly next to it. Obviously, we don't usually listen to instruments with our ears right up next to them. Close-miking can result not only in an unnatural-sounding magnification of detail, but also in tonal imbalances such as overly bright tones, sibilance, boominess, and the like.

In short, ambient stereo-miking tends to capture the natural characteristics of an instrument better than close mono-miking. Done well, it conveys localization (left-right placement in the stereo field) and, in comparison to mono, superior depth of field (the sense of distance from the instruments) and a fuller, more accurate "picture" of the recording space's acoustic properties (including reverberation, diffusion, and other elements).

Stereo-miking techniques can be broken down into three types, defined by how the microphones are positioned: coincident, near-coincident, and spaced. There are also a number of dedicated stereo microphones available that offer two or more diaphragms in the same mic body (for a variety of coincident-miking techniques), as well as stereo-miking systems such as the Crown SASS-P (Stereo Ambient Sampling System). This article, however, focuses on techniques that use individual microphones (usually two, but in some cases, three or more).

COINCIDENT As the name suggests, a coincident pair is two microphones positioned so that their capsules are as close to each other as possible. This closeness, or coincidence, ensures that the left and right signals have no phase differences and thus maintain their frequency response when summed to mono.

In coincident pairs, the stereo spread is created by amplitude alone. Directional microphones are most sensitive to sound coming from directly in front of the capsules (on axis) and are less sensitive to off-axis sound. In other words, the mic "hears" lots of sound coming from the front and much less from the sides. Therefore, sound sources positioned in the center of the angle between the two mics will appear in the center of the stereo field, whereas sources off to one side or the other will appear more on-axis to one microphone and thus will be reproduced louder on that side.

One of the simplest and most popular coincident-pair techniques is the XY pair. For balanced stereo reproduction, the two mics should be the same model and preferably matched (meaning that the manufacturer has taken extra care to ensure identical functioning). Also, it's important that the mics have directional polar patterns, whether cardioid, hypercardioid, or supercardioid.

The angle between the two capsules in an XY pair is usually set between 90 and 135 degrees (see Fig. 1). The wider the angle, the broader the stereo spread. However, compared with other stereo-miking techniques, even miking with a wide-angle XY pair will produce a rather narrow, focused stereo image generally lacking in depth and "airiness." In addition, there may be sound coloration due to the fact that off-axis sound constitutes a significant portion of pickup. (This can be a problem with some microphones and should be taken into account.)

When you're recording multiple instruments, you can widen the stereo spread by arranging the players in a semicircle around the XY pair. XY pairs also work well for close-miking instruments such as acoustic guitar and piano.

Naturally, the angle of the mics also affects the capture of reverberation and other ambient elements. An XY pair angled 90 degrees apart, for instance, places most of the room reverberation in the center; wider angles spread the reverberation across the stereo field.

Blumlein Pair A very effective variation on the coincident-cardioid pair is the Blumlein coincident pair, named after Alan Dower Blumlein, the pioneering electrical engineer and inventor who developed the technique. This setup requires two bidirectional mics (that is, mics with figure-8 polar patterns) whose capsules are angled 90 degrees apart. As with coincident-cardioid pairs, the sound source is centered between the two front-facing capsules (see Fig. 2). Essentially, the Blumlein array creates a four-sided polar pattern that gets summed to two channels. The left channel is made up of front-left and rear-right signals, and the right channel consists of front-right and rear-left signals.

To my ears, the stereo imaging captured by a Blumlein pair is hard to beat-the clarity and detail of stereo localization are simply nonpareil. (For information about stereo localization, see the sidebar "At Play in the Stereo Fields.") Another distinguishing characteristic is the uniform spread of room reverberation across the stereo field. Nonetheless, the use of Blumlein arrays does have some minor drawbacks.

One common problem is excessive room sound. To compensate, engineers often position the mics closer to the sound source, thereby increasing the ratio between direct and reverberant sound. This compensation, however, can overemphasize pickup from the center of the source, which in turn can make it harder to pinpoint the locations of sound sources at the far ends of the stereo spread (that is, at "the edges" of the 90-degree arc defined by the front-facing polar patterns). In other words, as you move the mics closer to the source, the stereo spread starts getting too wide.

Blumlein pairs also present the risk of unwanted phase anomalies, particularly when recording a sound source (for instance, a large ensemble) that stretches beyond 45 degrees from center. In such situations, phase anomalies are more likely to occur because the sound is being picked up by one mic's front diaphragm and the other mic's rear diaphragm, which are out of phase with one another. Again, the solution is to make sure that the sound source falls within the 90-degree arc defined by the front-facing microphone diaphragms.

For these reasons-especially if naturalness of sound is a primary concern (as it is when you're recording a string quartet, for example)-I wouldn't recommend using a Blumlein array in a highly reflective room or in a narrow room with a lot of sidewall reflection. On the other hand, if you want to capture lots of room sound (when recording a guitar amp, say), a Blumlein pair may be just what the doctor ordered.

Middle-Side Pair A middle-side (M-S) stereo pair employs one "mid" microphone (typically a directional mic with a cardioid pattern) that faces the sound source and one bidirectional "side" mic. The two mics are positioned so that the mid mic's primary axis is aligned with the side mic's null axis (the dead area on either side of the figure-8 pattern). This arrangement (see Fig. 3) captures a very stable center image and has the additional advantage of being thoroughly mono compatible.

To create a complete stereo image from M-S signals, you must "decode" them with a sum-and-difference matrix. The matrix, in a typical M-S setup, produces the left and right channels from the middle and side signals using the following equations (assuming the side mic's positive lobe is oriented to the left):

left channel = middle + side right channel = middle - side

One common way to decode M-S signals is to use three channels on a mixing board: the mid mic's signal is panned dead center in one channel and the figure-8 mic's signal is split (using a Y-cable) to feed two channels, one panned hard left and the other hard right (see Fig. 4). To complete the process, the polarity of one of the "side" channels is then reversed, either with a polarity-reversal switch on the console channel or with a special "polarity reversal" cable (one whose "hot" and "cold" wires have been reversed).

One potential problem with this form of sum-and-difference matrix, however, is that the side microphone is driving twice its normal load at only half its normal termination impedance-an imbalance that can result in distortion and a higher noise floor. For that reason, engineers who specialize in M-S recordings often employ gear with dedicated sum-and-difference-matrix capabilities. Units with active components are favored, as these provide a well-balanced matrix and, typically, a control system to vary the M-to-S ratio without damaging the signal-to-noise ratio, phase response, or frequency response. Many companies-for example, Calrec, E.M.T., Soundcraft, Telefunken, and Yamaha-have manufactured active M-S console modules. Currently available stand-alone models include Grace Designs' Lunatec V2 and Audio Engineering Associates' MS38 mkII, an outboard, active, line-level matrix control unit that provides a single M-to-S ratio control. Also on the market are stereo mics designed to record M-S and automatically decode the signals; for example, the Shure VP 88.

Part of the appeal of M-S recording is that you can manipulate the stereo image without having to rearrange the mics-even after the recording is completed. For example, you can tighten the stereo image and make the source seem closer by increasing the level of the mid-mic signal in relation to the side-mic signals. Conversely, increasing the level of the side-mic signals makes the source seem more distant.

The stereo image and overall sound can also be altered through panning: as the side-mic signals are panned closer to center, the stereo image gets narrower and the mid-mic signal gets louder, making the source seem closer. Panning the left and right side-mic channels to dead center (that is, to mono) effectively removes the side-mic signals and increases the mid-mic signal by 6 dB. The equation is (M + S) + (M - S) = 2M.

If a bidirectional mic is not available, you can construct a passable M-S setup by substituting a matched pair of cardioid microphones placed back-to-back (for the two side signals) with the polarity reversed on one of the mics. However, this method does not work as well as a true M-S setup. (For information about the origins of M-S recording, see the sidebar "A Walk on the Mid-Side.")

NEAR-COINCIDENT A near-coincident stereo array uses two directional microphones angled away from one another, with their grilles spaced a few inches apart in the horizontal plane (see Fig. 5). Again, the angle between the mics causes an intensity or level difference that creates the sense of the sound source's position in the stereo field. In addition, the space between the diaphragms introduces a time or phase difference to the stereo information, adding width and depth to the stereo spread and airiness to the recording. With near-coincident pairs, then, both amplitude and time difference contribute to stereo localization.

Results of near-coincident miking can vary considerably, depending on the frequency and polar-pattern (usually cardioid) characteristics of the particular microphones as well as on other factors. In general, watch out for too great an angle or space between the mics, as this will cause exaggerated separation. Similarly, too small an angle or space may result in an overly narrow stereo spread.

ORTF One of the most widely used and reliable near-coincident arrays is ORTF, so called because of its early use by the Office de Radiodiffusion Television Francaise. The ORTF technique specifies two cardioid mics with their capsules spaced 6.69 inches apart (the average distance between an adult's ears) and angled 110 degrees apart (see Fig. 6). Designed to mimic human hearing, ORTF provides a wide, accurate image and a good sense of depth. However, if the two signals are summed to mono, phase cancellations occur, which can cause comb-filtering effects that color the sound considerably. Yet it's the slight time delay between the left and right channels that contributes to the sense of stereo spread; if you don't need to reproduce the recording in mono, the ORTF technique will usually yield excellent results.

OSS The OSS (Optimal Stereo Signal) system was created in the 1980s by Swiss engineer Juerg Jecklin. The setup specifies a pair of omnidirectional microphones whose capsules are positioned 6.5 inches apart and separated by a sound-absorbing disk-called a Jecklin disk-measuring 11.81 inches in diameter. The microphone capsules are aligned with the center of the disk (see Fig. 7). Jecklin's original disk, made from an LP record, was covered on both sides with lamb's fleece, which is an excellent sound absorber.

Like ORTF, OSS is designed to simulate the positioning and pickup characteristics of human ears. The Jecklin disk takes OSS's simulation a step beyond ORTF's, providing separation and interference similar to that which is created by the human head. In large part, it is the addition of the disk that warrants the adjective optimal in Optimal Stereo Signal. When we listen to a sound source directly, our brains determine its location by analyzing the combination of three elements: sound intensity (level), time delays (phase relationships), and subtle variations in the way that different frequency ranges are picked up.

Near-coincident pairs employ only those first two elements to create stereo localization. Jecklin's OSS system, however, integrates all three elements to produce very lifelike stereo imaging. Note that OSS also utilizes omnidirectional, rather than unidirectional, microphones, which act much like our ears, picking up frequencies below 200 Hz equally. But as the frequencies increase (that is, as the pitches get higher), the disk (like a human head) blocks more sound, and the mics (like human ears) increasingly operate more directionally.

The OSS technique captures a full-bodied stereo spread rich in detail. However, the same qualities that allow "optimal" stereo signals make it impossible to sum those signals to mono without introducing phase distortion.

Building your own Jecklin disk is fairly easy (see Fig. 8 for a visual reference). Use wood, plastic, or a similarly stiff material (maybe a vinyl LP!) for the disk, which must be sturdy enough to serve as a mounting surface for the mic clips and mic stand. Cover both sides with at least 0.5 inches of acoustic foam. (I've also used several mouse pads glued together.) Mount the mic clips opposite each other near one edge of the disk so that the mic capsules are 6.5 inches apart and aligned with the center of the disk. Attach a mic-stand mount to the bottom of the disk, and you're ready to go.

Binaural Recording Binaural recording is clearly the stereo-recording technique that best approximates human hearing. Like OSS, it employs interference between the mic capsules-but in this case it's caused by a replica of a human head rather than by a disk (see the sidebar "Background on Binaural"). Binaural recordings are made using two tiny omnidirectional mics placed in the dummy head's ears (or, for even greater realism, using special mics placed in the ears of a human subject). The two signals are kept separate throughout the signal path from the mics to the corresponding left and right drivers of the listener's headphones. Other than the binaural source and a pair of stereo headphones, no special equipment is required to convey the binaural experience to the listener.

Conventional binaural recording is generally incompatible with stereo loudspeaker reproduction and can sound strange. However, some modern binaural setups employ equalization and special design features to deliver signals that sound good on stereo speakers. Even with the best setups, though, not all of the directional material translates to speakers; the eerie "you are there" binaural effect is largely lost, due to the leakage of sound cues intended for one ear into the other ear.

You can therefore best experience binaural recordings with headphones, through which the full 360-degree sphere of pickup can be heard. Sounds are then discernible from any location-right, left, front, back, up, and down-and their distance from the mics can also be clearly made out. The sense of realism is uncanny: a good rainstorm recorded binaurally can make you think that you're getting soaked.

Binaural recording works like this: the pinna, or outer ears, of the dummy head (or of the human subject wearing in-ear microphones) set up subtle interference patterns that give specific location to the sounds around the mics, thus positioning them accurately in space. Sounds coming from directly in front of the head, for example, bounce off the rear of the outer ear, sounds from below bounce off the top of the outer ear, and so forth. These ear-reflected sounds, in conjunction with sound that enters the ear canal directly, interrelate to create the interference patterns (known as Head Related Transfer Functions, or HRTFs) that the brain uses to determine sound location.

Binaural recordings are easily resolved to 5.1 and 7.1 (or any ".1," for that matter) surround sound and consequently are likely to become more prevalent in the audio-recording world. Binaural recording will also play a pivotal role in the development of virtual reality for computer games and other applications. Even for conventional stereo recording and playback, some engineers are of the opinion that modern binaural methods provide better-balanced and more-natural-sounding recordings than other stereo-miking techniques.

SPACED A spaced, or AB, stereo pair (see Fig. 9) consists of two identical microphones placed several feet apart and aimed directly at a sound source. The mics can have (or be set to) any polar pattern, but obviously the pattern must be the same for each mic. In addition, to maintain consistency of pickup, the mics are usually positioned parallel to one another. Not surprisingly, the stereo spread increases with the distance between the two mics.

Stereo localization in spaced pairs, unlike that in coincident and near-coincident arrays, is provided by time delay only. Sound from directly in front of the mics reaches both capsules simultaneously, so the recorded sound appears in the center of the stereo field. Off-center instruments, however, are closer to one microphone than to the other; therefore, the sound arrives first at the near mic. The brain uses the arrival-time differences to calculate spatial location.

One key to good spaced-pair recordings is finding the right balance between the stereo spread and the music as a whole. For example, to record a single instrument, the appropriate spread may be as little as 2 feet. To capture the full blend of sound from a large ensemble, on the other hand, you may need to place the microphones 12 or more feet apart.

The main drawback of spaced pairs is the diffuse pickup of off-center instruments. Therefore, it's sometimes a good idea to include a third microphone at the center point (especially when using widely spaced pairs) and mix the resulting signal to both channels. Then again, that same diffuse quality can lend a soft, desirable warmth to the natural reverberant qualities of the space. Your approach depends largely on what kind of sound you're after. Experiment with different mics as well as different spacings. If the mics are capturing too much room reverb, for example, try using directional rather than omnidirectional mics.

When positioning spaced pairs, it's helpful to heed the 3-to-1 rule, which states that the distance between the mics should be at least three times the distance between the mics and the source. (For more information, see "Recording Musician: Avoiding Phase Cancellation" in the July 1997 issue of EM.) In practice, though, following the 3-to-1 rule sometimes results in too large a "hole" in the center of the image, especially with widely spaced pairs aimed at large ensembles. Mic positioning is therefore critical. Ideally you should audition several setups, listen carefully, and compare the results to determine the optimal arrangement. Move the mics in small increments, because a relatively minor change can make a big difference in sound. Some phase cancellation is unavoidable, but when you find the optimal spacing between the mics, as well as the right balance of direct and reverberant sound, spaced pairs can create exceptional spaciousness and realism.

Decca Tree A useful variation on the spaced pair is the Decca Tree, which is essentially a spaced pair with a center mic for filling the hole in the center of the image. Engineers Arthur Haddy and Kenneth Wilkinson, among others, developed this array in 1954 while employed by the Decca Recording Company. In Haddy and Wilkinson's setup, the spaced-pair mics were placed 4.42 feet apart. The center mic was situated between and 2.17 feet in front of the spaced pair. At the time, Haddy and Wilkinson were using Neumann KM 56 nickel-diaphragm multipattern mics set to omni; later they changed to Neumann M 50 aluminum-diaphragm omnis.

The Decca Tree (see Fig. 10) was designed specifically for making stereo recordings of large orchestras. It usually consists of three omnidirectional microphones, although cardioids can also be used. Because the center mic sits in front of the spaced pair, the center signal in a Decca Tree is typically very robust. This can result in comb-filtering effects, so listen carefully to determine optimal placement.

For even larger orchestras, two additional microphones (making a total of five) are sometimes employed as "outriggers" on the sides of the "tree." These can be positioned anywhere from the outer edges of the sound source to approximately one-third of the hall's width in from the side of the hall. Each signal is mixed to the mic channel on its particular side. Decca Trees are generally mounted around 10 to 12 feet in the air and just behind the space where the conductor's head would be.

DOUBLE YOUR FUN I hope I've provided a useful survey of stereo-recording techniques and given you enough information to get you started on some stereo recordings of your own. Perhaps you have a recording project that could benefit from stereo-miking; you should now be able to analyze the situation and choose the most appropriate technique.

However, don't let the techniques limit you. Instead, experiment with them, try variations, and keep notes on the results-just in case you stumble onto something really good. Remember that thousands of talented sound engineers throughout the years have scratched their heads over countless strange acoustic dilemmas, eventually finding solutions through experimentation. Indeed, each of the methods described here was developed under just such circumstances. Who knows-maybe your experiments will lead to the development of a hot new stereo-recording technique.

Elizabeth Papapetrou has been recording music and writing for music magazines for 17 years. She loves microphones and specializes in recording solo and small-ensemble acoustic music. She is a performing singer, guitarist, and songwriter. Originally from the United Kingdom, she is now based in Gainesville, Florida. Special thanks to Bruce Bartlett of Crown International (www.crownaudio.com), Wes Dooley and Ron Streicher of Audio Engineering Associates (www.wesdooley.com), David Josephson of Josephson Engineering (www.josephson.com), and John Sunier (www.binaural.com).

Middle-side (M-S) stereo recording was developed for stereo broadcasts in the 1950s by Danish radio engineer Holger Lauridsen. A popular technique in Europe, it was relatively unknown in the United States until the advent of stereo TV broadcasts.

Engineers Wes Dooley and Ron Streicher popularized the technique in the United States with their widely read technical papers and their use of M-S in the 1980 Emmy Award-winning broadcast of Beethoven's Ninth Symphony (performed by the Los Angeles Philharmonic and the Los Angeles Master Chorale and broadcast by KCET, channel 28 in Los Angeles). M-S recording is now used by many diverse organizations, including National Public Radio, Louisiana State University, and cinema-sound facilities such as SounDeluxe and Lucasfilm's Sprocket Systems.

Stereo localization-defined here as the placement of sound sources in the recorded stereo field relative to their true positions during performance-varies in accuracy from one stereo-recording technique to the next. To compare the localization of the different techniques, stereo-miking expert Bruce Bartlett carried out a listening test using ten different stereo arrays (see On-Location Recording Techniques by Bruce Bartlett and Jenny Bartlett, Focal Press, 1999).

With each microphone array, Bartlett recorded speech sources at 0 degrees (center), 22.5 degrees (both left and right of center), and 45 degrees (also left and right). He did the test recordings twice: once in an anechoic chamber and again in a reverberant gymnasium. Afterward he played the results of each miking technique and asked his listeners to locate the sound sources' positions in the stereo field.

This table shows the averaged test results (the chamber and the gym recordings, by the way, were very similar). The upper portion shows the locations of the sound sources (labeled A through E). The bottom portion indicates where the listeners perceived them to be.

The first experiments with binaural principles were conducted in 1881 by Clement Ader at the Paris Exposition and Opera. Ader positioned a series of carbon microphone pairs (actually telephone elements) about 7 inches apart along the edge of the Paris Opera's stage. The opera performances were picked up by the mic pairs and sent on twin telephone lines to the few subscribers in France that had two lines installed in their homes. The earpiece of one line was held to the left ear, the earpiece of the other to the right. Ader said, "This 'double listening' to sound produces the same effects on the ear that the stereoscope produces on the eye."

Further experiments were sporadic over the next 60 years or so. In the mid-1920s, a few radio stations in the United States began broadcasting on two different frequencies, feeding each transmitter separately from a left-ear and right-ear mic in a dummy head in the studio (see Fig. A). Listeners were already using headsets, for the most part, so for this new type of broadcast they just used two radios, putting one mono headset (tuned to the left-ear station) to one ear and the other mono headset (tuned to the right-ear station) to the other.

By the 1950s, stereo and binaural techniques were often assumed to be the same. To clarify the terms, RCA Victor started including an explanation in the liner notes of all its 2-track stereo open-reel tape recordings. The notes read: "Stereophonic recording differs from Binaural (a term sometimes incorrectly applied to stereophonic records) in that the microphone placements are selected for loudspeaker reproduction. Binaural properly applies to a two-channel system designed for headphone reproduction. It thus requires the use of two channels fed by microphones spaced about seven inches apart (normal ear separation)."

Some West German radio stations devoted time to special binaural transmissions, often of radio dramas called Horspiele. Also, The Cabinet of Dr. Fritz series of binaural radio dramas from ZBS Productions was broadcast in the 1980s and early '90s by public radio stations here in the United States. Many of those same stations also carried binaural expert John Sunier's weekly program, Audiophile Audition, which featured his "All Binaural Special" broadcasts four times a year for more than 13 years.