Page  318 ï~~Studio Report Elektronisches Studio TU Berlin Technische UniversitAt Berlin Fachgebiet Kommunikationswissenschaft Folkmar Hein, Bernhard Feiten H51, Str. des 17. Juni 135 - D 10623 Berlin Tel.: +49-30-314 22821. FAX: +49-30-31421143 E-Mail:, Introduction When standing back and looking at the various subject matters which have been dealt with at our institute over the past 35 years, one suddenly becomes conscious of historical context. One recognises the slow movement from electrical engineering and electronics to our present-day digital technology. Problems encountered in the 60's have since become distant and almost incomprehensible. Often, proud results from the past will today be greeted only by amused chuckling - of course people in the year 2010 may also greet our present-day efforts in a similar manner. In looking back, however, we should not always expect to discover technical,,sensations" - especially not where art is concerned! Sensations are short-lived. Their importance quickly fades away and they remain with us only as vague images of the spirits of those past times. For this reason, we will be dealing here only with those research and development projects which were and remain artistically, musically or humanly relevant, and which - most importantly - have left their traces somewhere in musical thought. Certainly we must welcome research which corresponds to art; one which enables artists to communicate to technicians and vice versa. Research work done at our institute may be categorised according to various themes. Within undergraduate studies, as is customary in Germany, both a mid-study and a final diploma thesis are required; in post-graduate studies a single Masters' thesis is required (our department of communication science has been authorised to offer studies at the Masters level since 1979). From 1958 to 1985, before our department was able to offer official studies at the postgraduate level, we were concerned exclusively with works by students from other faculties (above all from the field of electrical engineering and somewhat later also from the field of computer science). Although communication science is a rather broad subject of study one can, from a practical viewpoint, categorise most work done within our department as being concerned with one of the following areas: electrical engineering, electronics, acoustics, computer science or musicology. In examining the approximately 300 thesis titles which have been presented as either mid-study, firstdiploma, masters or doctoral dissertations, it becomes apparent that in the beginning - that is to say in the 1950's - work was predominantly concerned with the development of technical equipment. Examination of thesis reference material also reveals information about the researchers themselves. As regards personnel, our situation has decidedly improved over the years: in 1970, five staff members were employed as compared to today's 14 permanent staff and 11 additional members who are involved in third-party projects. As a unifying thread through our almost four-decade history, one recognises the name of Dr. Manfred Krause who, following some years as teaching assistant, was appointed a full professorship at the institute in 1979. He followed Prof. Dr. Fritz Winckel who in 1954 had founded both the department and the studio. Upon appointment Dr Krause became responsible for all research and teaching at the institute and is himself engaged in the field of speech research and studio technology. In 1970 digital technology and therewith computer science entered into practical research work, and since that time it has step by step replaced practically all analogue techniques. Activity in the area of computer music was initiated at our institute in 1981 through the purchase of a Synclavier II, and was enhanced in 1984 by the acquisition of a VAX 11/780 computer. These acquisitions contributed to a considerable heightening of activity in software development and digital signal processing. Significantly enough, the first entry in our thesis index of 1958 reads:,,Frequency shifter (Siemens)" (additive shifter with two ring modulators). This entry arises repeatedly in our records - along with the ring modulator it gained a certain world-wide popularity among specialists. A few examples of activities will now be described in more detail and with historical references. These examples are considered of special general interest and have had, for the most part, artistic resonances. spatial sound - sound spaces The Osaka Project At the 1970 Worlds Fair in Osaka, the Federal Republic of Germany presented music to a world public in a specially-designed spherical pavilion. Following some preliminary differences of opinion, agreement was reached to perform predominantly contemporary music. Even as late as the summer of 1969 original concepts for the pavilion underwent changes, but the pavilions basic architectural design - which had been drawn up by Fritz Bornemann and based on suggestions by Karlheinz Stockhausen for a spherical hall - remained in tact. It was decided however, that along with the music of Stockhausen, other contemporary German composers also should be heard. Erhard GroBkopf, Eberhard Schoner, Gerd Zacher, Bernd Alois Zimmermann and Boris Blacher were commissioned works. The initial planning of technical aspects by the Siemens company, which provided for an eight-chan 4B.3 318 ICMC Proceedings 1993

Page  319 ï~~nel mixer controlling eight rings of loudspeakers, was taken over by a team from the TU Berlin under the direction of Prof. Boris Blacher, Prof. Fritz Winckel, Dipl.Ing. Manfred Krause and tonmeister Rudiger Rifer. Their work consisted of completing the sound system through the addition of controls for electroacoustic sound diffusion, allowing for the movement of sounds through space. The already-completed eightchannel mixer remained unchanged. The TUJ team had already gained experience with four-channel sound projection through the earlier 1966 performance of Blacher's opera,,Zwischenflllle bei einer Notlandung" at the Hamburgischen Staatsoper. Firstly, the very reverberant acoustics desired by Stockhausen had to be avoided - otherwise the clear localisation of moving sound sources would be near to impossible, and there also would arise the danger of echoes and the effects of focal points. A multi-layered, lightly-built sound absorption construction was designed and realised for the space. The reverberation time was reduced to about 2.2 seconds. For sound spatialization control, a 7 x 7 matrix was conceived to distribute through electronicallyprogrammable junctions to a total of 49 loudspeaker groups on seven reduced mixer channels. The eighth channel was reserved to control the matrix. The two 1-inch-four-channel tape recorders, which had been planned on by Stockhausen, were replaced by two mechanically-linked 4-channel-35mm perforated-magnetic-film machines, allowing for the synchronised playback of the seven music tracks and one control track. The control track contained in itself 14 controlsignal channels: seven for controlling the loudspeakers (coded with amplitude modulation) and seven for programming the brightness of 49 groups of lights (coded with frequency modulation). In one of the institute's conference halls a provisional playback system was installed in order to arrive at least at an approximate impression of the spatialization effects. In Berlin it was possible to record sound movements onto the control track. It would also be possible though, to record spatial movements oneself live at the hall in Osaka. For this purpose, special input equipment was developed including,,sensor balls" on which control buttons corresponded to specific loudspeaker positions, as well as the socalled,,Rotationsmiihle" (rotation mill) designed to create circular movements in space. The project's barely sufficient preparation time of about nine months was cause for hectic activity, but thanks to special efforts from the electrical engineer in charge of technical implementation - Dipl.Ing. Claus Amberg - the system was shipped and installed in Osaka according to schedule. The scheduling of music at the pavilion was organised so that each morning the commissioned compositions would be presented in a rhythm of 15-minutes; afternoons were reserved for the works of Stockhausen. The technical set-up worked essentially problem-free for the onehalf-year duration of the EXPO, although in the rainy season the lower part of the hail stood underwater and amplifiers had to be raised up on wooden constructions. Sound-space research The desire to create sound spaces and spatial sounds in realtime is common goal. Unfortunately, nearly all projects of this sort have been limited to calculations of intensity, and thereby have generally yielded unsatisfactory results - primarily due to underestimating the importance of time-related localisation. Non-realtime calculations, on the other hand, which allow one to take all possible parameters into consideration, yield better and easily achievable results. Such systems are employed at our institute (for example in Cmusic). Following the Osaka project, the institute did not return to making developments in the area of,,soundspace control" until 1985 - except for the realisation of a,,movement simulator" which was developed in 1982. Along with intensity calculations this simulator could produce rather convincing Doppler-shift effects. The diploma thesis of Erwin Schmidt concerned a computer-controlled 4 x 24 matrix based on a multiplying DAC. This graphically-supported application permitted the generation and recall of various sound figures which, through sequencing and overlapping, could then be replayed as a sound-space score. Because switching noises were not entirely unavoidable, this concept was abandoned. The other option, namely the use of VCAs, was subsequently investigated. After finding support from outside the university, research in this area - headed by Werner Schaller - led to the creation of a mature, stand-alone system named the RKS. Also, parallel to the research and development of this comfortable yet expensive solution, another system was being developed by Thorsten Radtke. This second system had only four channels of sound output but could be produced at much lower cost. The system is easily controlled by MIDI data and although it involves MDACs, it excludes the previously problematic switching noises (with the use of zero axis crossing detection). For this,,MIDI mixer", which distributes up to 15 sources to four loudspeakers, Thomas Seelig developed a graphic environment for the ATARI. Here, spatial paths for the different 15 sources are simply drawn on the screen; paths may be edited and are synchronously distributed to the speakers while being shown as moving points on the screen. A further sound space project headed by Werner Schaller concerns eidophonic as well as orthophonic processes. With eidophonie as according to Scherer, a fast sampling of the surroundings with the aid of electronically-controlled direction characteristics is produced in a special stereo intensity microphone; subsequently decoed signals can be reproduced over any number of loudspeakers. This process does not however seem to deliver satisfying results. Orthophonie on the other hand employs a four-membrane microphone for first-order applications (one with omnidirectional and three with perpendicularly-standing figureeight characteristics). As is also possible and common with stereo technology, any three-dimensional direction can theoretically be characterised by the weighting and addition/subtraction of the four microphone signals. A decided advantage of this arrange ICMC Proceedings 1993 319 4B.3

Page  320 ï~~ment lies in the fact that the number of speakers used for replay does not need to be determined until the time of reproduction, and that replay is possible over any number of speakers from 6 to 32. For better, higher-order orthophonic results necessary microphones also must be developed, and this is a goal of the research as well. Orthophonic signals - even higher-order ones - also can be calculated from monophonic sources. Research in this particular area of orthophonie could lead to a new manipulation technology for the localisation and spatial placement of sounds, and therewith to new developments in virtual spatial movements and spatialization effects (like reverb) for electroacoustic music. An orthophonic mixer also is imaginable. Sound visualisation Even as early as 1947, so-called,,visible speech" processes existed. Our institute possessed one of the first sonagrams (from the Key Electric company) and employed this not only for speech research but also in the research of musical and various other sounds.. Sonagrams deliver quasi three-dimensional images of sound with the horizontal axis representing time, the vertical axis representing frequency and degree of darkness representing the amplitude of sound. For certain applications this form of representation is inappropriate or awkward; a second type of solution seemed to be required: a running on-screen representation of sound in realtime. So came into being in 1979 the first version of the,,sonascope" - a diploma work from Hubertus Becker. An improved version of this real time application, realised with the use of analogue filters, was delivered to the Heinrich-Strobel-Stiftung of the SWF (Southwest Radio), and eventually became known for its many years of faithful service. The sonascope had at its disposal 21 third-of-an-octave filters in a band ranging from 70 to 3500 hertz. The viewable window (as with sonagrams) of 2.5 seconds, was made up of 128 samples from the various filter envelopes. Once the input signal surpassed a setable input threshold the image moved from right to left across the screen. Darkness as characterising intensity was represented in 4-bit grey scale (2.5 dB per step). The main disadvantage of the equipment was said to be a certain temperature-related instability and its rather cumbersome filter adjustment. The use of this technique - namely rendering sound and its temporal structures easily readable - seemed so convincing that in 1987 with the aid of new computer technology and through the development of speciallydesigned hardware and software, work on the project was taken up again. However, this time the framework of the project was not to be concerned with the visualisation of music but rather with that of speech - specifically as a training aid for the deaf. System requirements related to this application in the area of speech analysis were: preemphasis; an analysis band ranging from 50 to 7000 Hz; 10-bit resolution; 8ms segmentation; the realisation of variable analysisbandwidths according to frequency groups; special considerations concerning differentiation in represent ing vowels, consonants, fricatives and nasals in contrast to the representation of impulses. Special attention also had to be paid to the recognition of fundamental frequencies as well as to the exact representation of their temporal characteristics. This implied the recognition of both voiced and non voiced qualities, and necessitated the analysis of fundamentals with a high-resolution FFT. The system's constant and moving pattern generation in a quasi-sonagraphic form required inclusion of the following elements: continuous changes in articulation had to be continuously drawn on the screen; the segment properties of speech had to shown; speech dynamics had to be represented by colour intervals related to researches in subjective experience. The project is sponsored by BMFT (DFG). A final example concerned with the visualisation of sound falls into the area of computer music. In considering the sonagram as a pure sound analysis technique (that is to say the representation of points of time-frequency-amplitude) one might also imagine reversing the process so that graphic sound analysis would result in sound resynthesis (in other words an additive synthesis based on the previously done analysis). This process has already been known for some time as the "Vocoder" - though originally, the graphic representation of analysis data was not a matter of concern. Over the past four years a program package which essentially takes up and realises this idea, has been developed under the direction of Dr. Bernhard Feiten. The system is based on a Micro VAX 3600 computer running under the ULTRIX operating system as well as on DEC Workstations under X-Window. The analysis program,,spare" (spectral analysis reduction) - implemented and extended by Holger Becker and based on work by McAulay, Quatieri, Smith and Serra - incorporates properties of harmonic structures and hearing into data reduction (for example the incorporation of such considerations as sound masking). The objective is to approximate in line segments, the temporal characteristics of a sound's component tones. For this purpose peak detection and tracking is employed; interpolations between analysis blocks allows the creation of various line segments and tracks. The well-known problem of having to trade off between higher frequency-resolution on the one hand, and higher time-resolution on the other hand, has been reduced here through the implementation of an adaptive control of the effective width of the analysis window in the Fourier transformation. A variable control of hopsize and the number of tracks to be generated resulted in better and more efficient implementation. The reduction of data flow lies between 1:3 and 1:80. Output is ASCII-text or binary code. Sound analysis results from spare may be further worked on and edited in a second program called,,sounded". Sounded was developed by Frank HeimNicher for a Workstation under X-Window. The program produces a temporal graphic representation consisting of points (showing where partial tones change) and tracks (lines between these points). The y-axis can be customize to show the cent-, hertz- or 4B.3 320 ICMC Proceedings 1993

Page  321 ï~~logarithmic amplitude scale; the functions copy, paste and move have not yet implemented. New points and tracks may be added by the user at any time. Of course one also may realise a purely synthetic score without the initial use of analysis data at all. Individual points can be numerically edited by hand or displaced through,,dragging" with the mouse. Results of an editing session can subsequently be synthesised with the,,adsy" program. In reference to this subject the reader is referred to further CARL-system developments, for example a graphic four-channel 32-track mixing editor (xmix4); a soundfile editor (sofEd); a channel-vocoder (crossyn). Other Activities Firstly, in connection with research at our institute, certain cooperations with other institutions and research establishments should be mentioned. A contract of cooperation between the TU and Hochschule der Kflnste specifies our involvement in the formation of HdK composition students. A contract with the DAAD specifies the accessibility of our facilities to DAAD guests in exchange for which there is financial compensation as well as the right for us to award a four-month stipend for work in the studio. Aspects of research also enter into consideration in our planning of artistic activities. At the INVENTIONEN festival - as well as at other international events presented under our direction - there have been numerous symposiums, workshops and publications with research-oriented content. In this manner Berlin also has become familiar on a first hand basis with the music and activities of other studios such as INA.GRM, IRCAM, EMS, STEIM, CARL, Les Atelier Upic, Studio Basel, ZKM. Among publications - apart from the INVENTIONEN program books which themselves contain a wide variety of information and essays on music research - is the project and publication entitled the "Documentation of Electroacoustic Music in Europe". The book bearing this title contains data about European studios concerned with the production or research and teaching of electroacoustic music, as well as data about works which have been produced or conceived in these studios. The documentation contains two differently sorted Work Lists, and a Studio List with the mailing addresses, telephone, FAX and email numbers of 261 European studios (including 31 in Italy), and with information about studio equipment and additional special notes. The Studio Work List includes 5,991 productions from the 52 more major studios. The Composers Work List contains all 8,157 of the documented works, including those works from private studios and "independent" composers which do not appear in the studio work list. An index referencing studios and composers, as well as a list of abbreviations used are found in the appendix. The documentation can be ordered through our studio. (trans. Robin Minard) Equipment (excerpts) Mixer: Speakers Taperecorder Soundcraft S8000 (36 / 8 / 2), Case. 2 Hill-Mix (16/4/2) four K&H 092; four 086, sounddiffusion: d&b four F1, two B1 1-inch M15A 8-tr; 1-inch M10-8-tr (optional 4-tr-PB-head). 1/2-inch M15-4-tr 1/4-inch: two MI5A 2-tr; two Studer A80R; TEAC A-3440 4-tr; others Noise-Reduction 20 TelCom, 8 Dolby-A, 4 dBx-. Zeta 3 -Synchronizer two PCM 701 with two Umatic, two Beta, Beta-HiFi, SVHS HiFi RDAT: two Panasonic SV-3700, transportabler Sony TCD-Dl0, AIWA Excelia Effects Publison IF 90, 5.1 sec; 2 Drawmer M500; 2 Klark DN 410; EMT 445 (21 sec),; Microphones Neumann-Mikros (eg. SM 69, 2 M269c; RSM 190i, KU 81i), two Sennheiser MKE Analog synthesizer SynLab (4 VCA, 5 VCO, 4 ADSR, 2 VCF, 2 RM, 2 S&H, 2 noiseG&F, 2 Env.Follower). 4 DC-Prxlzisions Spannungs-Geber, 16 MIDI to 12-Bit-DAC Digital synthesizer Synclavier 11(16 voices, Sample to disk, VT 100_GB, 20 MB Disk) MIDI-devices two Atari 1040ST; Akai S 1000 (18 MB); Roland A-SO; Yamaha DX100; MIDI-Mixer, 16 MIDI-DAC's 12 Bit; YES-audio MIDI-Fader. Cooper CS-10 MI)I-Projection-system MD12 (12 MIDI-controlled Kodak-Caroussel) MIDI Timepiece II; JamBox 4+; 2 Apple - MIDI - Interfaces. Roland CP 40 Macintoshs Quadra 950 (36/230 MB), 3 GB ext., ProTools 16 ch, Centris 650 (20/105 MB), 1 GB ext, ProTools 8 ch. 11 fx (8/105 MB); SEl/40 (4/40 MB); two 40 MB ProDrive, CD-ROM. Personal LaserWriter NTR; Image Writer II Mac-Software ProTools, Sound Designer II & DNR; StudioVISION, Finale 2.6, Cubase Audio, Turbosynth, MAX 2.5, Csound, Word 5.1, FileMakerPro 2.0, MacX, MacUNIX, Timbuktu, Sonagramm (von KTH); FileForce; Csound-881; soundhack; reverb DEC-VAX Micro-VAX 3600, two RA 82 (je 600 MB), RA 81 (450 MB), Lineprinter LP 26 DSC-200 with four DACs, two ADCs, Graphikprinter, 3 VT220, 5 VT240, DEC-Station, two VAX-Stations 3100, VAX-Station 2000 VAX-Software Ultrix 32; internet; Fortran F77; C; XWindow, TUB-software sonapr, sounded, spare, adsy, sofed, xmix, crossyn IRCAM-Software Chant und MOSAIC CARL-Software csound, cmusic, Player, phasevocoder ICMC Proceedings 1993 321 4B.3