(back to SpatialLocalizationPaper)

Our treatment of StereoRecording concluded by highlighting factors that hinder completely accurate playback: channel crosstalk and room reflections. Thankfully for most hi-fi owners, engineers take these factors into account when recording and mixing down their productions. Spot mics accentuate solos that might be heard in the balcony but not on stage where it competes with the rest of the ensemble; room mics add depth if a large-scale effect is desired; even minimalists work to place their stereo pair at a reasonable distance.* What if we don't want to compromise? ?RTF introduced a promising direction to target our focus: rather than fiddle endlessly with pan-pots, let physics do the work for us.

Binaural recording takes this premise to its logical extreme. Microphones are placed in the ear of either a human subject or "dummy head," with the resulting recording played back over headphones. In this way, it bears a conceptual similarity to the 3D stereographs popular before the rise of film. The technique is used by concert-goers to record with stealthy electret condensors, by psychoacoustic researchers refining the transfer functions and other aspects of the aural system, by audiologists examining a patient or testing a new hearing aid, by architects investigating a concert hall, and by a select few commercial producers (mostly in Germany).

Compared with typical headphone reproduction, this method is a huge win. Traditional stereo recordings rely on physical channel crosstalk to help separate the stereo image; on headphones this creates the illusion of three "blobs" lateralized at the ears and in the middle of the head, unless special circuitry is employed to recreate stereo crosstalk electronically. Even then, the effect is still one of lateralization, far removed from the authentic spatial reproduction of binaural recording. Binaural recordings, on the other hand, create a vast three-dimensional soundstage where sounds are localized outside the head, since the head-related transfer function (HRTF) applied by the microphone placement corresponds nearly identically to the effect a listener's head and pinnae and so forth would provide were she in the concert hall.

How does it compare with a well-mastered stereo recording on a good hi-fi? The effects can only be described as very different. In any case, speaker and headphone technology generally do not compete against each other, so the more pressing question is compatibility. Unfortunately, early** attempts at "Kunstkopf stereophony" failed in this regard: sounds entering a dummy head are equalized much as they would by a human head, but when played back over loudspeakers are equalized again upon reaching the ear, leading to unnatural tone quality. They were also of generally poor quality, unable to distinguish front/rear directionality nor to rival conventional mics for self-noise (Blauert, 359-360).

Completely accurate reproduction of binaural recordings over speakers requires special circuitry at the listener's end.*** Reducing the goal to merely a soundstage comparable to that of a stereo recording, however, invites a solution: filter the signal from the microphones in such a way as to achieve loudspeaker compatibility, then when using headphones add another filter in series that restores the recorded HRTFs (taking into account the equalization of headphones) (Blauert, 360-2). Needless to say this would never catch on.

Luckily the ability to use high-quality microphones, in addition to lowering the noise floor to a level comparable to normal recordings, enables the equalization of the transducer to a diffuse field. In practice inside a dummy head, this creates a large gap (-15dB) below 10 KHz. Such recordings provide acceptable localization and tone color over loudspeakers, although they require diffuse-field equalized headphone systems. Since the publication Blauert's 2nd edition, it turns out that the audiophile world has released scores of diffuse-field headphones -- after all, as Blauert noted, they actually sound more pleasing on conventionally-mic'd signals too -- as well as popularized the use of small circuits to adjust the equalization (among other things, crossfeed also being popular), notably a large adjustable notch filter at about 7.5 KHz high and -15dB strong. Not coincidentally, these are to be tweaked to the half-wavelength resonance between one's eardrum and the earpiece.

Back on the recording side, the Kunstkopf concept has the backing of a major microphone manufacturer in Neumann, its main competitor being a much smaller company from Aachen. Do-it-yourself dummy head projects abound, with good results (though generally not very stereo-compatible) being obtained even with relatively simple constructions of baffles, mic holders, and duct-taped minidisc recorders. The crafty electrets-posing-as-earbuds method mentioned above also remains popular.

*The major exception are room modes. Every room has at least 3 corresponding to height, width, and depth that (1) can produce strong resonances that unlike "normal" reverb directly interfere with music enjoyment (2) vary between every room, and thus cannot be engineered at production time. If two modes coincide the problem is that much worse. Since this section is not about stereo reproduction, I'll just chalk this footnote up as another reason to prefer binaural technology.

**Early is relative here -- 1970s, let's say. The first use of binaural technology was actually at the 1881 Paris Opera via dual telephone lines, well predating traditional stereophony.

***Even then, to recreate the spatial impression of a binaural recording, elaborate physical setups are needed. One brief fad in the 1970s had audiophiles place a large baffle between the left and right speakers running from the back of the room right up to the listening position.

Disclaimer: the author has spent way too much money on headphones, headphone amps, crossfeed circuitry, and binaural recordings

onward to 3DSynthesis