by DENIS VAUGHAN
[Denis Vaughan is the Musical Director of the State Opera of South Australia in Adelaide.]
For many years, the ability to locate a sound's origin was thought to be solely due to loudness or intensity; a softer sound seems farther away. As applied to recording, a slight increase in the volume of one channel will shift an image in the direction of the louder channel, and if both channels are raised simultaneously, the image will seem nearer to us.
Later on, it was realized that relative timing of the two channels could be used to affect location of the recorded stereo image. The incredible speed of the brain allows us to perceive time differences as short as only 0.007 mS, which is equivalent to the sound source moving just 1° to the side. In my own monitoring of recordings, I am able to place an oboe, for example, within 1° or 2° of a fixed spot; the mechanics of this common feat requires rather infinitesimal perceptions by the brain.
In addition, the ear and brain can tell us whether the origin of the sound is above or below us, and in general, stereo technology has taken little account of this fact. There are fail-safe mechanisms in various parts of the body, and I would therefore dare not say that our perception of a sound's elevation is based exclusively on a single system. However, as far as I can ascertain, the chief system seems to be associated with our perception of timbre. The recent data on the transformation characteristics of the outer ear by Mehrgardt and Mellert [1], shown in the figures, plot how the outer ear functions as a filter between a constant level sound field and the inner ear.
These changes are made in everything that we hear, and so we are used to making allowances for them. This is why we often use a weighting filter in making a measurement.
Fig. 1--How the ear canal filters sound frequencies en route to the eardrum,
as measured by various researchers. (After Mehrgardt and Mellert.)
Fig. 2--How the ear's response to different frequencies varies with the sound
source's lateral (azimuth) angle to the right ear, from 0° (in front of the
listener) through 180° (behind listener); positive angles are on right side
of head, negative angles on left.
(After Mehrgardt and Mellert; dashed lines show earlier results by Shaw.)
Figures 2 through 4 show how the perception of various frequencies of constant level vary with angle. The first fact to be drawn from these graphs from Mehrgardt and Mellert is that we hear a perceptibly different timbre at every angle of reception. These timbre changes are clarified, I hope, by the respective Tables. (In order to interpret these readings properly, we must not forget that the ear canal imposes another curve on all our listening, which compensates for some of the extremes in the curves listed above; see Fig. 1.) However, only when we analyze the Tables in detail is it possible to see that there is always some frequency for any two adjacent angles of reception that fluctuates quite substantially--sufficient to help us identify one angle's timbre from its neighbor's. The quickest test of the veracity of this is to rub your fingers lightly together, and then move them in a circular motion about one ear, while still rubbing. It is always possible to tell where they are, without any help from the other ear which, if plugged, does not perceive the signal.
A further look at the comparative timbre readings also shows that at no angle can we hear anything resembling "flat" response. We can, therefore, ask, "Is there any point in creating a flat response if no one can hear it?" To facilitate quick study, the Tables have dark shading over any reading +5 dB or above; Table I has a light tone over any negative quantity under -5 dB, and Tables II and Ill have a light tone over all negative quantities.
This throws into prompt relief the extremes of each timbre, which are never less than 15 dB apart and sometimes as much as 21.5 dB (at 1801. What is surprising is the extent to which the mind registers and discriminates between all these fluctuating curves, sufficient to tell us whether we are being attacked from directly above or from, say, 47° elevation on one side.
Before examining some of the mechanisms involved, it is worth looking at a fundamental finding which Blauert described in his 1969 article, "Sound Localization in the Median Plane" (Acustica, Vol. 22, pg. 211). He writes, "The direction of sound sensation can be altered merely by altering the spectrum of the sound signal that reaches the eardrum, at least when visual cues are not available. If the spectrum at the eardrum is constant, the sound incidence does not play any role for the direction of the sound sensation. To determine a certain direction of the sound sensation, the sound signal must obviously be linearly distorted in such a way as is normally done by the head and pinna, when the sound wave reaches the observer from that very direction from which the sound sensation is desired to appear."
There are some frequencies which our brain tends to localize in one direction only, and they tend not to change from person to person as much as external ears (pinnas) appear to change.
A very rough schematic summary is given in Fig. 5, which is taken from Blauert's book, Raumliches Horen [2]. It is maddening to tune a frequency generator to 8 kHz, only to find that no matter where the speakers are placed, or how one turns one's head, the sound is always overhead. This effect disappears for me quite shortly above and below this frequency. However, it does indicate that there are certain stable points of reference. The extent to which that frequency, 8 kHz, is filtered into or out of our reception by the pinna, helps our brain identify the extent to which the arriving sound is coming from above (or at least appears to be). A further tendency in all listeners is one which Roffler and Butler reported [3]; specifically, that out of 50 subjects, almost all identified higher frequencies as coming from higher angles. In their test, no one pointed to the loudspeaker actually emitting the sound but, to a very common degree, at the angles depicted in Fig. 6. I assume that no one ever pointed to the lowest speaker, because no sounds below 250 Hz were used in the test. The angles were not fixed, because when the test subjects were placed further from the speakers, the angles decreased in nearly all cases, which suggests that visual stimuli also play a part. However, the general identification of high frequencies with higher elevation still persists.
Fig. 3--Frequency response of the ear at different azimuth angles from 0°
(straight ahead) through 90° (on the side of the ear being measured) to 180°
(behind the head). (After Mehrgardt and Mellen.)
Fig. 4--How the ear's response to different frequencies varies with the vertical
angle (elevation) of the sound source, from 0° (ear level) to 180° (above the
listener's head). (After Mehrgardt and Mellert.)
above: Adelaide Festival Theatre, Adelaide, South Australia [Architects should
get away from the tyranny of long reverberation times when they design acoustical
spaces. ]
Fig. 5--Human interpretation of the direction of sound sources at various
frequencies. (After Roffler and Butler.)
Fig. 6--Identification of elevation with frequency. (After Roffler and Butler.)
Table I--How the outer ear's response varies at different frequencies and
sound-source azimuth angles.
Heavy shading indicates areas where response rises by 5 dB or more; light shading indicates where response dips by -5 dB or more. (After Mehrgardt and Mellert.)
To me, as an orchestra conductor, the collation of this material illuminates a number of hitherto total mysteries, as well as helping to explain why I obtain great satisfaction from certain musical and aural stimuli. The feeling of being truly surrounded by sound, if there are sound sources or strong reflections from sufficient angles, will base its satisfaction on at least three elements: (1) A much wider frequency range will be audible. Frequencies which had been filtered out by the pinna at one angle will stand a chance of being transferred louder at another angle; (2) All the network of my direction-finding faculties will be involved, so that my sensory participation in the sound is greatly increased, and (3) The arrival times of so many impulses will perforce be scattered, and the more impulses I receive, the richer the sound will seem to be. (The further one progresses from the simultaneity of mono, the better!) A conductor worth his salt could thus distinguish between the arrival times of the sound of each of 20 violinists, all of them sitting at least 4 mS from each other. This is a huge time, at least in comparison with the 0.007-mS minimum audible time difference mentioned near the beginning of this article.
However, the fact of the wider and more complete frequency range, heard only when the sound comes from all angles, arouses some serious new thoughts about the reproduction of music. For example, it appears that we hear ultra-high frequencies better when they come from the side! What qualities do we need to hear most in music? Well, it seems to me that architects should get away from the tyranny of trying to get long reverberation times when they design spaces for acoustical uses. This has caused several overly large halls to be built in the last few years. In addition, I have drawn up a list of musical qualities in sound, and I've given them a certain aesthetic order, along with a brief description of their physical counterparts. It reads thus:
Richness: Powerful multiple reflections from all angles shortly after the arrival of the original sound.
Density: A large number of reflections (impulses) in a short interval of time.
Warmth: A bass-heavy frequency response curve, with a peak between 125 and 250 Hz, and a smooth falling away above that.
Intimacy: An adequate number of lateral reflections of frequencies above 10 kHz, which are absorbed noticeably by the atmosphere after travelling only 15 meters.
Clarity: An adequate number of early lateral reflections, so that the upper frequencies are adequately presented, and an absence of confusion due to the predominance of reverb rather than first reflections.
Singing Tone: A peak in the reverb curve, usually with a rise-time of about 75 to 100 mS, a plateau of 50 mS, and then a smooth decay starting 150 mS after the original sound.
I have shown this list to many recording producers, technicians, conductors, and acousticians, and I am gratified to note that most of them agree that richness is the most desirable quality. Jordan pointed out in his recent book on concert halls and theaters that satisfactory acoustics had been obtained with reverberation times which ranged from 1.4 to 2.8 seconds, so that obviously should not be the controlling factor. In an article for a British publication [4], I listed a number of speculations about how multi source reproduction of sound might be developed comparatively inexpensively, to give a stereo picture which no longer required the listener to remain in a fixed position in relation to the placement of the speakers.
But other fields are also waiting to be explored. If a headphone eliminates the functioning of the pinna, can some signals be fed to it which, using Blauert's system, restore a fully detailed stereo picture, one with the frontal elements firmly fixed? The frontal spectrum in my lists is so widely divergent from the rear spectrum, that there must be a way to simulate its position clearly by electrical means. Can we hoist various instruments up and down by increasing or decreasing the amount of 8 kHz in their reproduction? At this point I'd like to recount one of my own experiences which I feel will be useful to all readers; it does not always coincide with the recommendations from loudspeaker manufacturers.
I have found that if I place stereo speakers at least 4 feet from an end wall, the resultant reflection from that far wall helps the image come to life, as it allows an effect to take place which cannot be achieved with speakers set in, on, or near the wall. In a reverberant room, with my Tannoys 6 feet from the rear wall, I was able to recreate the impression that the orchestra sat behind the speakers at exactly the same distance as the real orchestra actually sat behind the microphones during the recording session. In this way, for me, the speakers disappeared entirely in the final image.
I therefore feel the timing of the first frontal reflection in a room may be decisive to the fixation of a three-dimensional stereo image, if the timbre fluctuations outlined above can be incorporated into the reproduction.
Denis Vaughan--To me, as a conductor, this material illuminates
some mysteries, and helps explain my satisfaction from certain aural stimuli.
Table II--The data of Table I, adjusted to show in-pinna lateral response
relative to frontal response at the same frequencies. Here, light shading indicates
all areas where response is lower than response to sounds dead ahead; dark
shading indicates where response rises by 5 dB or more.
Table III--How the ear's response varies with frequency and the sound source's
vertical angle, relative to its response to sounds straight ahead. Heavy shading
indicates response rises of 5 dB or more; light shading indicates response
dips.
References
1. Mehrgardt, S. and V. Mellert, "Transformation Characteristics of the External Human Ear," J.A.S.A., Vol. 61, No. 6, pg. 1567.
2. Blauert, J., Ráumliches Hóren bei einer Schallquelle, Hirzel-Verlag, Stuttgart, 1974.
3. Roffler, S. K. and Robert A. Butler, "Localization of Tonal Stimuli in the Vertical Plane," J.A.S.A., Vol. 43, No. 6, pg. 1260, and "Factors that Influence the Localization of Sound in the Vertical Plane," J.A.S.A., Vol. 43, No. 6, pg. 1255.
4. Vaughan, Denis, Musical Times, Jan., Feb., Mar. 1981.
5. Jordan, Wilhelm, Acoustical Design of Concert Halls and Theatres, Applied Science Publishers, London, 1980.
6. Shaw, E. A. G., "Transformation of Sound Pressure Level from the Free Field to the Eardrum in the Horizontal Plane," J.A.S.A., Vol. 56, No. 6, pg. 1848.
(Audio magazine, Dec. 1983)
Also see:
Doug Sax: On Current Record Technology and Music Systems of the Future (Mar. 1980)
Sound Reinforcement for the Amateur (Mar. 1984)
= = = =