Gabriel Weinreich - weinreic@umich.edu
Randall Laboratory of Physics
University of Michigan
Ann Arbor, MI 48109-1120
Popular version of paper 2pEA4
Presented Tuesday afternoon, 17 June 1997
133rd Meeting of ASA, State College, Pennsylvania
Embargoed until 17 June 1997
If the sound of a live solo violin is compared with that of the same instrument played back through a high-grade sound reproduction system, the loss of perceived quality is enormous. This is not due to the electronic system's deficiencies in frequency response, distortion, etc., but comes from the DIRECTIONAL PATTERN in which a loudspeaker projects its sound. In fact, a moment's thought will reveal that, regardless of the nature of the original source (piano, violin, voice, or full orchestra), by the time the sound emanates from a loudspeaker its directional properties will have been steamrolled into those of the loudspeaker. In the words of the famous composer and conductor Pierre Boulez, "The loudspeaker anonymizes the real source."
Why should the directional pattern make any difference in what we hear? After all, isn't the only important signal the one that is directed straight toward our ear? The answer is that we almost always listen to music in an enclosed area, rather than in empty space. Our brain then has at its disposal not only the direct signals from the two separate ears, but also the subtle time delays and phase shifts that arise from wall reflections. This, together with information that we unconsciously gather by moving our heads around during a concert, enriches our perception with a profusion of information arising ultimately from the directional pattern of the source.
The projected sound of an ordinary loudspeaker is generally characterized by a concentration of intensity in the forward ("axial") direction at high frequencies. The higher the frequency, the stronger is this concentration; but the transition happens slowly and relatively smoothly. This behavior is very similar to what we would hear if the original source of sound were in an adjoining room, and the separating wall contained a circular hole of the same size as the loudspeaker.
By contrast, the directional pattern of a violin is completely different: at frequencies above about 850 hertz, it varies very rapidly both with angle and with frequency. This means that (a) at any one frequency the sound emerges in fairly sharp directional "beacons"; (b) the orientation and strength of these "beacons" changes completely when the frequency is altered by as little as one semitone (about 5%).
When we listen to a solo violin, we perceive those beacons which are not aimed at our ears in terms of their reflections from the walls on which they impinge (this sometimes causes us to identify various notes as coming from different directions). A violin thereby acquires an "acoustic size" much larger than its physical size, endowing its music with a characteristic spatial sense which a normal loudspeaker does not have.
We have named this special property of certain musical instruments - the property of having a directional pattern which varies rapidly both with angle and with frequency - "Directional Tone Color," or "DTC." For the case of a violin, the theoretical and experimental description of DTC, along with various musical implications, was discussed in a recent article [Gabriel Weinreich, "Directional tone color," J. Acoust. Soc. Am. 101(4), April 1997, pp. 2338-2346].
In today's presentation, we describe and demonstrate a new type of speaker system called "DTC Loudspeaker," whose purpose is to endow an electronic sound source with the characteristic of "directional tone color" which it normally lacks. In its latest embodiment (the "Mark VII"), our system consists of a single woofer and four tweeters. Although the whole system is driven by only one monophonic channel, the signal for each of the tweeters is electronically processed to introduce rapid phase and amplitude variations as a function of frequency. The phenomenon of wave interference among those four sources then produces a directional "antenna pattern" which varies sharply both with angle and with frequency; in other words, it exhibits "directional tone color." As we already saw, such behavior causes the brain's normal directional cues to be sufficiently confused to make it seem on occasion as though each note were coming from a different direction; in other words, the source of the music appears to be much more extended in space than it really is.
Although the DTC speaker was originally developed using the solo violin as prototype, it soon became clear that its applicability is much wider than that. It is, for example, particularly successful for pipe organ music, since in this case every note of the original music really does come from a different direction. As part of today's talk, we shall demonstrate this effect, which also takes place with orchestral music and any other sounds that come from extended sources.
An interesting counter-case occurs for a solo singer, whose voice tends to lose clarity when played through the DTC speaker. This is to be expected theoretically, since the opening of the human mouth is small compared to most of the wavelengths that are important in the musical spectrum. For this reason, there can be no appreciable interference among different parts of the source, and therefore no strong directional variations in the radiated intensity.
All our experiments so far have been done with monophonic signals (we get our program material from standard CDs simply by throwing away one of the channels). The question then naturally arises as to the exact role played by traditional stereo, and whether something might be gained by combining the techniques of stereo and DTC. In comparing the two, the following points should be noted:
(a) A traditional stereo setup tends to be especially effective for one place in the room, whereas a DTC speaker, like the original instrument, sounds equally good regardless of the listening point.
(b) A traditional stereo setup is useful for LOCALIZING the source of sound, whereas the DTC speaker can rather be said to DELOCALIZE it. In many, if not most, musical applications the latter appears to be more effective than the former, for reasons that we have already indicated.
(c) In many cases, there is some support for the point of view that, in spite of the supposed "theoretical" basis for stereo, it actually functions for the most part as a "poor man's DTC."
(d) As a point of objective comparison, it is interesting to remark that a traditional stereo setup is NON-HOLOGRAPHIC, in the sense that either channel alone reproduces the sound of the source (though, of course, the stereo effect is then lost). By contrast, the DTC system is HOLOGRAPHIC, in that any one of the four tweeters by itself does not constitute a pleasant reproduction.
In fact, of course, the question is not whether to choose stereo over DTC or vice versa, but how best to combine the two. To answer it, as well as other questions arising from this work, much more research will need to be done.