(back to SpatialLocalizationPaper)

Of interest in this age of recorded music is finding a way to convey a musical presentation without the aid of a concert stage. The classic Bell Labs experiment established a 3-channel speaker array as optimal for faux localization (in the horizontal plane anyway), but stereophony is what caught on. We'll examine a few common methods of acheiving a pleasing stereo image using what we know about the psychoacoustics of localization.

The simplest, and yet most common method of recording I refer to as "pan-potted mono." Individual instruments or voices are recorded in a studio, with all effects, gain, and so on added in parallel; they are "arranged" in the final mix by directly (often logarithmically) weighting the amount of signal each channel gets. This system provides decent localization of auditory events between the speakers, using the blurring characteristic of hearing equivalent sources with different intensities. However, should the listener move more than a foot or two from the equadistant line between the speakers, the precedence effect will apply, and all the sounds will appear to come from one speaker! (Moore, 164) This disadvantage obviously applies to every stereo system, but characteristic to this recording technique is the lack of additional positional cues to aid the listener.*

The other major class of techniques I call "live recording." These can range from minimalist recordings where only a stereo pair of mics is used to setups with dozens of spot mics, with the defining characteristic being that the entire ensemble -- and nearly as importantly, the room -- is recorded at once. Of course, merely placing mics in a reverberent room does not in itself generate the additional temporal cues we are looking for. Sure enough, basic stereo techniques such as XY merely use the directionality of a mic pair to spread the image across the channels. Unlike studio-tracked recordings they can create the illusion of soundstages wider and deeper than the speaker placements, but the precedence effect still dominates. There are countless techniques that attempt to refine these principles, but I'd like to cover ?RTF:

Named for the Austrian radio organization that standardized it, it separates directional mics by 17 cm and angles them at 110 degrees. With the capsules apart, off-center sounds thus generate signals that are slightly out of phase. There is some math to justify the exact result of cardiod patterns at this spacing & angle -- and even then, it is of course a function of frequency -- but in the end it provides the ear with some phase differences that mimick those heard by a person in the original room.** Applying some of the processing discussed in BinauralLocalization, then, we obtain some positional cues that improve the quality of the spatial reproduction, although the precedence effect is not completely ameliorated. After all, 17 cm is conspicuously close to the width of a human head; ears are generally omnidirectional, but 110 degrees somewhat reflects the rear occlusion provided by the pinnae. Needless to say the transfer function of music processed by such an arrangement is not a terribly accurate reproduction of the aural system, but it represents a good compromise.

*Of course, it has several advantages that make it extremely popular. Moore names most of them on p. 237: ability to equalize and otherwise affect each instrument; to record multiple takes with ease; to use electronic/computer instruments; etc.

**This is not to say that stereo-pair techniques allow one to place mics where the audience would sit. When added to the stereo crosstalk and room reflections in the listener's home, the total amount of reverb would muddy the sound.

onward to BinauralRecording