The basic idea of dummy head technology or head-related stereophony is true-to-original reproduction of sound events. Dummy head technology’s great promise is to store an entire sound event, including its spatial characteristics, enabling the listener to later (re-)locate sounds in space as if being in the original recording situation. This quest for facsimile sound reproduction could easily be read as one (tiny) episode in media history’s (apparent) progress towards greater definition, fidelity and truthfulness. However, dummy head technology was, and still is, not only a sound reproduction technology but also an important measuring instrument for studying human spatial hearing. In my paper, I will follow the KU80 dummy head through different research laboratories and broadcasting stations. I will investigate the main actors’ struggle for interpretative authority, attempts to conceptualise the phenomenon of human spatial hearing and efforts to improve binaural technology accordingly. To address these issues I will distinguish three modes of hearing tests deployed in dummy head research: measurements, correlations, and consumption.*

Bell Labs are often credited for building the first dummy head: named Oscar. Oscar was part of research efforts to improve the telephone system. Based on the observation that listening with two ears provides higher fidelity, Oscar was used in a series of tests with the Philadelphia Symphony Orchestra during winter/spring of 1931/32. In a memorandum Bell engineers K. Hammer and W. Snow noted that Oscar “will reproduce satisfactorily the intensity and frequency range of ordinary speech, music and most noises and gives a fair degree of localization.” However, one problem most listeners experienced was “that all sounds came from the rear.” This front-rear inversion was not at all surprising, as we will learn later, because the two microphones were mounted on Oscars’ cheeks and not inside his ears.[i]

bell_localization

The “Localization Phenomenon” (Source: Hammer/Snow 1932).

Another well known dummy head was built a few years later by de Boer and Vermeulen, engineers at the Philips Laboratory in Eindhoven. They first used a sphere with two microphones to construct a hearing aid. Their concept was based on the observation that a person with impaired hearing could better follow a conversation between more than two people if the transmitted acoustic signals would contain some spatial information (de Boer/Vermeulen 1939). Without knowing the exact effect de Boer and Vermeulen tried to use the so called ‘cocktail party’ effect: the ability of human spatial hearing to focus auditory attention on a particular stimulus and filter out a range of other stimuli. De Boer and Vermeulen assumed that spatial hearing was based on time and sound pressure differences between the two ear signals, but they still could not explain all localization phenomena. To investigate the role of the human head in spatial hearing they also built a manikin with microphones mounted in the ears. They called their artificial head “Kunstkopf”, a term that was later also used in English. De Boer and Vermeulen compared the manikin with their spherical microphone, but they could not find any differences – which is, compared to today’s knowledge, rather surprising.

These early examples show the still limited knowledge of human spatial hearing. They further illustrate that dummy head technology was well situated between scientific laboratory studies and the development of commercial sound technologies. The Neumann dummy head KU80 is another case of a binaural technology that was situated between these two poles.

Inventing Dummy Head Technology

In 1967, Georg Plenge, Ralf Kürer, and Henning Wilkens, three engineers at the Heinrich-Hertz Institute in Berlin, started their investigation of dummy head technology. They were looking for a new method to evaluate the acoustic quality of concert halls. Initially, they used more traditional methods like the measurement of impulse responses. However, with this technique they could describe the temporal extent of an acoustic event inside a given room, but they could not tell anything about the spatial characteristics of this acoustic event (Plenge 1971). The assessment of these spatial characteristics was reserved to subjective listening tests. Their new plan was to use dummy head recordings to better control these listening tests: The expected facsimile recordings of an “acoustic event” at a given place in a concert hall should be reproduced in the laboratory and allow any number of participants (when listening through headphones!) to experience an “auditory event” as being in the original recording situation. Plenge, Kürer, and Wilkens found some information in the literature about recent dummy head experiments, and in the Spring of 1969, they built their own dummy head. Their general assumption was that spatial hearing depended on the filtering effect of a human’s upper torso, head and outer ears (consisting of the pinna, the ear canal and the eardrum), thus they needed a careful replication of a human head and the outer ears. With the help of a make-up artist and a plastic surgeon, they produced plaster copies and mounted two studio microphones (Neumann KM83) at the end of the ear canal. A special coupler was supposed to reproduce the eardrum impedance.[ii] A proof of concept was done by Henning Wilkens in his diploma project. Then, Ralf Kürer used this dummy head for a larger investigation of the Berlin philharmonic hall. In his PhD thesis he showed that dummy head recordings could indeed be used to find correlations between impulse response measurements and subjective judgements of acoustic quality (Kürer 1972).

kuerer_dummyhead

First experimental dummy head of Kürer/Wilkens/Plenge (Source: Private Collection Ralf Kürer).

For a second test campaign more dummy heads were built (Wilkens 1975). The new aim was to test the acoustic quality of different places in several concert halls; for this purpose parallel binaural recordings of orchestra music had to be made. After some test recordings in different Berlin venues, Georg Plenge, Henning Wilkens and two assistants got the chance to accompany the Berlin Philharmonic Orchestra with Herbert von Karajan on a concert tour through West-Germany. Henning Wilkens used the dummy head recordings made during this trip in his PhD thesis. He developed a questionnaire with 15 contrasting pairs. Then, he had 40 test subjects listen to the facsimile recordings of the same piece of music made at different seats in one concert hall and asked them to judge the acoustic quality of these different places. Like Kürer, Wilkens could show interesting correlations between subjective judgements and measured physical properties.[iii]

However, Plenge, Kürer, and Wilkens not only found some nice correlations between subjective and objective measurements, they also realized that they could use their dummy heads to produce music recordings with astonishing spatial characteristics and good overall sound quality. They filed a patent and settled an agreement with the Berlin based microphone company Neumann to produce dummy head microphones: later named KU80. In parallel, they contacted different broadcasting stations to see if they would be interested in testing dummy head technology. Much to their chagrin, producers and recording engineers turned down this offer. It was not until 1973 that the Berlin radio station RIAS approached the dummy head inventors. After listening to some test recordings, members of the radio play division decided to produce the first binaural radio play – based on the science-fiction novel The Demolished Man by Alfred Bester. It was presented at the Berlin broadcasting fair the same year and simultaneously broadcasted by RIAS. Reactions of trade fair audiences and radio listeners were overwhelmingly positive—also most newspapers and consumer magazines praised dummy head technology as super stereo providing the true-to-original spatial impression of the original recording situation. The reactions of recording experts were much less enthusiastic: they criticized the poor reproduction of high frequencies and the mandatory use of headphones. Unfortunately, I can not go into this intriguing conflict between recording professionals, dummy head supporters and radio listeners, but I will come back to the question of headphone use and the demand for loudspeaker compatibility.

Improving Dummy Head Technology – Studying Spatial Hearing

First, we will go back to the research labs: The studies of Plenge, Kürer and Wilkens (and related work by Damaske and Wagener at the III. Institute of Physics in Göttingen) attracted attention by other researchers, and between 1970 and the early 1980s numerous investigations of dummy head technology and the entangled question of human spatial hearing were launched. A close reading of trade journals and archival sources from different labs reveals two entangled research strands: the first aimed at improving binaural technology, the second aimed at understanding human spatial hearing. The common point of departure was a set of localization problems of current binaural systems: in particular the problems of in-head-localization, front-rear inversions, and the elevation of auditory events. These three localization problems were equally experienced by many radio listeners and most test persons (Kuhl/Plantz 1975).

For the improvement of binaural technology researchers chose two different and rivalling approaches. The first approach was called the “physicalistic method” (Blauert 1978). It followed the conviction that “seen from the perspective of electrical communication engineering the principle of head-related stereophony can only consist in the undistorted measurement, transmission and reproduction of sound pressure signals at the human eardrums.” (Platte cited in Conrad 1976). The second approach took recent psychoacoustical findings into consideration, and stated that “transformations of sound signals between exit and entrance [of a binaural system] can be ignored, if they have no effect on the auditory experience (auditory event), i.e. if they are inaudible.” (Blauert 1974: 222). Proponents of each approach entailed different experimental settings and testing practices. Still, both groups used similar bricolage techniques to study and improve binaural technology: i.e. they slightly modified existing technologies to study the measurable or audible effects of each alteration.

Engineers at the RWTH Aachen University favoured the “physicalist method”; they developed new techniques to measure the so called head-related transfer functions inside the human ear, to then compare measuring results with dummy head signals. Head-related transfer functions describe the complex filtering of sound events through head and outer ear. Here you can see the monaural transfer functions for identical sound events coming from four different directions. Researchers in Aachen developed a device to insert tiny probe microphones into the ear canal of test persons to measure transfer functions as close as possible to the eardrum plane (e.g. Platte/Laws 1978). The received measurements implied that the replication of the human eardrum impedance was decisive for correct sound localizations (Theile 1981a). One obstacle was that the exact eardrum impedance of living human subjects could not be taken; thus researchers had to rely on approximations (Platte 1978; Laws 1978; Hudde 1978).

probe_microphone

Measurements are taken inside a subject’s ear (Source: Blauert 1969/1970).

The psychoacoustic approach made use of the finding that an “auditory event”, like sound localization, did not depend on the true-to-original reproduction of an “acoustic event”;[iv] the reproduction of decisive spatial characteristics was sufficient. Applying this knowledge, researchers at the Ruhr-University Bochum tried, for example, to find the “typical test person” through a combination of acoustic measurements with subjective listening tests. The “typical person” was the person whose head-related transfer functions are closest to the mean head-related transfer function of all test subjects. Researchers assumed that “listening” through a replication of this person’s pinna would allow most test subjects to make accurate sound localizations.

By the end of the 1970s head-related stereo had become the standard method to investigate true-to-original sound reproduction (Blauert 1978); and psychoacoustic experiments with binaural recordings helped to better understand the phenomenon of spatial hearing – the second research strand (Gottlob 1978). In particular new knowledge about monaural and binaural head-related transfer functions was acquired. Based on these new findings, researchers in Bochum could e.g. show that the KU80’s poor reproduction of frequencies above 7 kilohertz was responsible for localisation failures in the median plane (Hudde/Schröter 1981; Wollherr 1981).

Still, some findings had a rather short shelf-life and revealed the still limited understanding of human spatial hearing. As mentioned before, several research teams confirmed the great significance of eardrum impedance, but in 1980 Herbert Hudde (Hagen) and Jürgen Schröter (Bochum) (1981) could prove that ear canal and eardrum impedance had no measurable impact on sound localization. Hudde and Schröter were surprised by their findings, because one of them had also emphasized the opposite assumption before. As an explanation they mentioned a basic dilemma of electrical communication engineering: the construction of acoustical devices, as dummy head microphones, highly depended on measurable properties, but in the end such technologies were supposed to elicit “auditory events” which could not be easily captured in numbers.

This dilemma seemed to be ingrained into the laboratory practice of electrical communication engineering. Most experiments were designed to investigate selected aspects of “spatial hearing”, they aimed at formulating functional models for isolated relations between acoustic events and auditory events. For instance, one case examining the phenomenon of directional hearing formulated a functional model that was only valid for narrow-band loudspeaker signals, in the free sound field, for the horizontal plane (Wendt 1963). In 1981, Günther Theile, head of the department for sound recording and reproduction at the Institute for Broadcasting Technology in Munich, fiercely criticized this naïve approach: “How can you justify to study directional hearing separated from distance hearing?” he tauntingly asked. He emphasized that both direction and distance are entangled spatial coordinates of one “auditory event”. He continued: “A functional model of directional hearing is not a model of perception, it does not necessarily describe a specific function of the ear in the case of spatial hearing.” (Theile 1981b: 158) Instead, Theile argued for a holistic approach and formulated a new theory of sound localization, based on recent findings in neurophysiology and perceptual research., He called it the “association model” (ibid.). As you can see in this figure the association model consists of two consecutive decoding processes: first “the decoding of the spatial information is dealt with in the localisation association stage”, the second higher stage decodes the Gestalt information of an acoustic event. Both processes are based on associative pattern recognition. As a consequence of his theoretical model, Theile argued that only diffuse field correction of dummy head microphones and headphones could provide suitable ear signals that would allow correct sound localization. Here, he was in opposition with most of his colleagues who had recently argued for free field correction of dummy head microphones. This preference for free field correction was driven by their desire to use dummy head microphones in acoustic research. Here free field corrected microphones were standard for measurements in anechoic environments.[v] However, Theile’s preference was also driven by a desire for compatibility, but in his case compatibility with ordinary studio microphones – usually diffuse field corrected.

Research for Commercialising Dummy Head Stereo

This brings me to the third mode of binaural hearing tests: consumption. Today, I will focus on research projects that aimed at improving the compatibility of binaural technology with radio broadcasting requirements.

Georg Plenge (1978) formulated three conditions for the smooth introduction of dummy head technology in radio broadcasting. First, it had to be compatible with existing studio and broadcasting equipment; second, it had not to interfere with existing technical and administrative procedures; third, it had to be compatible with regular radio reception. The first two conditions were (technically) not obstacles, because dummy head stereo was compatible with established two channel recording and transmission technology; but the third condition posed a serious problem. Radio listeners had to rely on headphones to experience the “three-dimensional” sound of dummy head stereo. But stereo receivers with headphones were not at all standard equipment for ordinary radio listeners. In 1974, the Bavarian Broadcasting Service estimated that a mere 2% of radio listeners could receive dummy head stereo that way.[vi] This is why loudspeaker compatibility became an important research question.

The first approach to tackle this issue was ambitious: researchers were looking for technical solutions to preserve the spatial impression of dummy head stereo when reproduced with normal stereo loudspeakers. P. Damaske and B. Wagener from the III. Institute of Physics in Göttingen mixed dummy head stereo signals with compensation signals to make sure that the left loudspeaker signal reached only the left ear and the right loudspeaker only the right ear. This method worked quite well, but only in an anechoic environment and the precondition that listeners stayed fix in the median plane between the two speakers, thus it could only be used in laboratory experiments (Damaske/Mellert 1969/70). Plenge, Kürer and Wilkens developed another method with four loudspeakers. The idea was the same, here the two rear speakers would send compensation signals to prevent cross talk of dummy head signals (Kürer/Plenge/Wilkens 1973; Kuhl/Plantz 1975). The advantage was that this method did not require an anechoic environment, however the listener still had to stay within a rather small range, and the overall technical effort was too large for commercial success.

four_loudspeakers

Set-up for reproducing binaural stereophony with four laudspeakers (Source: Kuhl/Plantz 1975).

The second approach followed a different reading of Plenge’s third condition: now dummy head stereo had to be compatible with standard mono and stereo loudspeaker reception but without providing spatial impressions. Still, the overall sound quality had to be of comparable quality with conventional mono and stereo recordings. According to listening tests with recording professionals from different broadcasting stations, Neumann’s KU80 did not meet this criterion: its tone quality was not satisfactory when reproduced with standard loudspeakers. The technical explanation was a dip in the higher frequency range. This problem was solved with the improved model KU81 (developed 1979/1980). New assessments with expert listeners had ambiguous results: many listeners confirmed the better tone quality of KU81, and its sonic loudspeaker compatibility – measurements of its frequency response confirmed these judgments. However, another group of recording professionals continued to criticize dummy head stereo as not being on eye (or better ear) height with conventional stereo recordings. The problem with these listeners was that they were implicitly judging dummy head technology with existing market conditions in mind. Since the introduction of multi-track recording and close microphoning, new preferences for recorded music had emerged (see Schmidt-Horning 2013). Dummy head technology was in principle not attuned with this new style (in popular and classical) music recording: binaural recordings could not be mixed with other microphone signals without destroying the spatial illusion and thus offered less transparency and clarity than conventional recordings with numerous spot microphones. A test report of the BBC research laboratory noted that dummy head compatibility was indeed not a question of technical but artistic judgement.[vii]

Conclusion

To conclude, the development of working dummy head microphones fuelled research in the understanding of human spatial hearing during the 1970s. The improvement of dummy head technology and the study of spatial hearing were intimately linked to each other in most projects. Different research groups struggled for interpretational authority over explaining the phenomenon of sound localization, and achieving true-to-original sound reproduction. In their investigations, researchers deployed three different modes of spatial hearing tests: the first mode used comparisons of (objective) measurements of dummy head signals and human ear signals – with congruent signals indicating true-to-original sound reproduction; the second mode searched for correlations between measurements and subjective listening tests – this approach acknowledged psychoacoustic findings that even identical signals would not necessarily cause congruent auditory events; the third mode was based on expert judgements of exemplary music recordings – this mode took into account that commercially successful dummy head stereo had to comply with current production and consumption preferences.

The different disciplinary backgrounds can help to explain preferences for one of these three testing modes: electrical communication engineers preferred the “physicalist method” of objective measurements and aimed at formulating functional models for different sound localization phenomena; communication engineers (with an additional education as Tonmeister)[viii] and psychoacousticians relied on correlations between objective and subjective measurements to understand human spatial hearing. However, the physicalist or the correlation mode was also chosen with a hidden agenda in mind: the former mode was useful to turn dummy head technology into a standard microphone that complied with existing acoustic standards; the latter helped to develop dummy head technology into a studio microphone technically compatible with state-of-the-art recording and broadcasting technology. Finally, engineers involved in the commercialisation of dummy head stereo preferred “spatial hearing tests” with expert listeners. This method did not comply with any technical standards but with the commercial values of the recording business.

Literature

Blauert, Jens (1974). Vergleich unterschiedlicher Systeme zur originalgetreuen elektroakustischen Übertragung. Rundfunktechnische Mitteilungen 18 (4): 222-227.

Blauert, J. (1978). Vorbemerkung. Rundfunktechnische Mitteilungen 22 (4): 195-196.

de Boer, K. & R. Vermeulen (1939). Eine Anlage für einen Schwerhörigen. Philips’ Technische Rundschau 4 (11): 329-332.

Conrad, J. (1976). Kunstkopf-Stereofonie und Vorne-Ortung. Funkschau 47 (3): 95-96.

Damaske, P. & V. Mellert (1969/70). Ein Verfahren zur richtungstreuen Schallabbildung des oberen Halbraums über zwei Lautsprecher. Acustica 22: 153-162.

Gottlob, D. (1978). Anwendung der kopfbezogenen Stereofonie in der akustischen Forschung. Rundfunktechnische Mitteilungen 22 (4): 214-216.

Hudde, H. (1978). Methoden zur Bestimmung der menschlichen Trommelfellimpedanz unter Berücksichtigung der Querschnittsfunktion des Ohrkanals. Rundfunktechnische Mitteilungen 22 (4): 206-208.

Hudde, H. & J. Schröter (1981). Verbesserungen am Neumann-Kunstkopf. Rundfunktechnische Mitteilungen 25 (1): 1-6.

Kürer, R. (1972). Untersuchungen zur Auswertung von Impulsmessungen in der Raumakustik. PhD thesis TU Berlin.

Kürer, R., G. Plenge & H. Wilkens (1973). Wiedergabe von Kunstkopfsignalen über Lautsprecher. radio mentor electronic 39 (11): 512-514.

Kuhl, W. & R. Plantz (1975). Kopfbezogene Stereophonie und andere Arten der Schallübertragung im Vergleich mit dem natürlichen Hören. Rundfunktechnische Mitteilungen 19 (3): 120-132.

Laws, P. (1978). Messung und Nachbildung von Trommelfellimpedanzen. Rundfunktechnische Mitteilungen 22 (4): 201-206.

Platte, H.-J. (1978) Probleme bei Messung und Nachbildung von Aussenohrübertragungsfunktionen. Rundfunktechnische Mitteilungen 22 (4):198-201.

Platte, H.-J. & P. Laws (1978). Technische Probleme beim Einsatz kopfbezogener Stereofoner Übertragungsverfahren. Rundfunktechnische Mitteilungen 22 (1): 22-27.

Plenge, G. (1971). Über die Hörbarkeit kleiner Änderungen der Impulsantwort eines Raumes. Acustica 25: 315-325.

Plenge, G. (1978). Probleme bei der Einführung der Kunstkopfstereofonie im Hörrundfunk. Rundfunktechnische Mitteilungen 22 (4): 216-218.

Schmidt-Horning, S. (2013). Chasing Sound. Technology, culture & the Art of Studio Recording from Edison to the LP. Baltimore: Johns Hopkins University Press.

Theile, G. (1981a). Zur Kompatibilität von Kunstkopfsignalen mit Intensitätsstereofonen Signalen bei Lautsprecherwiedergabe: die Klangfarbe. Rundfunktechnische Mitteilungen 25 (4): 146-154.

Theile, G. (1981b). Zur Theorie der optimalen Wiedergabe von stereofonen Signalen über Lautsprecher und Kopfhörer. Rundfunktechnische Mitteilungen 25 (4): 155-170.

Wendt. K. (1963). Das Richtungshören bei der Überlagerung zweier Schallfelder bei Intensitäts- und Laufzeitstereophonie. PhD thesis TH Aachen.

Wilkens, H. (1975). Mehrdimensionale Beschreibung subjektiver Beurteilungen der Akustik von Konzertsälen. PhD thesis TU Berlin.

Wollherr, H. (1981). Mikrofonankopplung an das Aussenohr eines neuen Kunstkopfes. Rundfunktechnische Mitteilungen 25 (4): 141-145.

Notes

* Paper presented at the workshop “Testing Hearing: Science, Art, Industry”, Max-Planck Institute for the History of Science, December 4-5, 2015

[i] Karl Hammer & W. Snow, Memorandum MM-3950, 2.12.1932, Study of Speech and Hearing at Bell Telephone Laboratories, CD-ROM compiled by Christine Ronkovic & Jont Allen.

[ii] Henning Wilkens, Erinnerungen aus dem beruflichen Leben, 6.12.2006, Private Collection.

[iii] Damaske and Wagener conducted similar experiments at the III. Institute of Physics, University of Göttingen.

[iv] This distinction was introduced by Jens Blauert (1970) Ein Versuch zum Richtungshören bei gleichzeitiger optischer Stimulation. Acustica 23: 118-119.

[v] According to DIN 45580, standard headphones were also free field corrected.

[vi] Aktenvermerk Filbig v. 18.12.1974, Signatur 12543, BR Historical Archive.

[vii] D. J. Meares und E. W. Taylor, BBC Research Department Technical Memorandum No. PH-1739. An Assessment of the Neumann Artificial Head Type KU81, File Gerhard Spikofski, Dummy Head Collection, Archive Deutsches Museum Munich.

[viii] Electrical communication engineers at the TU Berlin could choose Tonmeister programme as part of their study.

Advertisements