Computers and Live Electronic Music: Some Solutions, Many ProblemsSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact firstname.lastname@example.org to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 135 ï~~COMPUTERS AND LIVE ELECTRONIC MIUSIC: SOME SOLUTIONS,P MANY ROBLEMS Simon Emmerson Music Department City University Northampton Square London EClV OHB Abstract: In the 1980s two types of computer composition emerged from the studio to the performance space, one more interested in event processing, the other in signal processing. Although recent systems suggest this divide can be ended, the legacy remains. This paper examines the current state of gesture transducers and their influence on human/computer interaction in live performance and concludes with a discussion of modes of use of the computer within the live performance set-up or 'who controls whom or what?'. The paper will be presented from the composer's and performer's point of view aiming to clarify questions, to establish limitations of existing equipment and to make demands for future developments from hardware and software engineers. Event processing versus signal processing Sometimes words lead us down paths of assumptions. An 'event' presupposes a change in some perceived quality at a clearly identifiable time; hence an event cannot take place without a 'signal'; yet somehow we have been lead to separate the two - the 'score' from the 'orchestra'. The development of literacy in the Western Art Music tradition has lead to a progressive separation of the parameters of music production. An historical sequence is also a hierarchy of decreasing accuracy of notation: pitch, time, dynamic are the primary parameters of the Western score while timbre and expression - small systematic modifications of the previous - remain recent developments within notation only vaguely defined and kept largely separate. The development of the score tradition follows an increasing tendency to define 'performance practice'. Scores are strictly dynamic not absolute: just as we rewrite the scores of Guillaume de Machault a future century may rewrite those of Beethoven or Boulez to include more performance practice information. Now if we reexamine the word 'event' it could be a change in any of the above parameters; but it is generally used only of the primary three. We thus see the artificial divide between 'event processing' and 'signal processing' which has resulted effectively in the development of two types of computer music composition. This divide has been perpetuated in all the MUSIC series of computer music languages since the 1950s and has been enshrined both in the MIDI protocol and in the controllability of most commercial synthesis systems to date. For studio composers this divide may be overcome by sheer hard work 'against' the system as appropriate - the means of achieving a desired sound combination may be tried and retried. For the live performance composer/performer however there is literally not the time! There appears to have emerged types of 'software gesture' which parallel 'instrumental gesture' (as limited by the physical structure of the instrument and the human body). This is often related to the path of least resistance while learning the software. Examples include 'cut and paste' procedures and systematic manipulations of motivic and thematic material, rhythmic cell manipulations etc. This leads to the perpetuation of musical ideas based on abstract (sometimes complex) event procedures but yet often realised into sound using simple aural resources (which have tended to be 'pre-set' and remain relatively unmanipulated). The musical interest is generated largely through 'event change' procedures. ICMC 135
Page 136 ï~~In the converse ('spectro-morphological') approach to electroacoustic music, in which 'signal change' procedures dominate, the paradox is that the relational or combinatorial aspect of 'events' is relatively simple. Of course, the signals they trigger are another matter! These may have been the result of months of phase-vocoding, filtering, combination etc.. The manipulation of these complex sonic entities is often separated completely from the 'event' processing - perhaps effected on a completely different system. But recent software developments suggest we are now in a position to incorporate signal processing instructions into the 'score'. We are forced to look at 'events' and 'controls' in hierarchies combined into local structures - very different from the linear view encouraged by previously available software. Properly manipulated, such event complexes become effectively complex objects (signals) in their own right. We must therefore rethink ideas such as 'rates of eventuation' and dispense with notions of 'combinatorial richness' (inherited from serialism and other quasialgorithmic methodologies). We need to replace this with a more sophisticated hierarchy in which 'events' control all levels of musical argument from the micro-timbral to the formal but yet are accessible and controllable in parallel. The move onto the concert platform History repeats itself. In the 1960s composers were able to move many of the studio techniques developed over the previous decade onto the concert platform through miniaturisation and with the addition of more comprehensive manual and pedal controls available following the advent of voltage controlled devices. Similarly, the 1980s saw a great increase in interest in 'live' interactive systems (for example as reported at the Steim Symposia of 1985 and 1986) based on miniature versions of systems developed in the previous decade and also adding a new generation of transducers for control. But unlike in the earlier decade this was coupled with an accentuation on complex event structures and combinatorial rule and decision taking protocols at the expense of complex sound objects. To draw a parallel with linguistics, this reflects too much of an interest in syntax with no corresponding research into possible phonology. Our second type of composition was rarely addressed live on the concert platform. It is evident that commercial manufacturers were caught out in the first generation of digital sound processors by the incredibly obvious desire of composers and performers to vary the process parameters in real time. Now I want to argue for a difficult distinction. 'Real time' is not that same as 'live'. At the ICMC in Glasgow (September 1990), one had the spectacle of four performers sitting giving serious attention to VDUs, and rather in the manner of an advanced weapons interception system 'reacting' to emanations from each other. 'Live' it was only in the sense that it was not pre-recorded - the actual actions of the performers were almost exclusively finger on QWERTY keyboard, even the few times MIDI keyboards were used the actions often had the same trigger function, Thus 'instrumental' and 'human' gesture were replaced by algorithm, the human relegated to the level of a sophisticated trigger/response mechanism. This description is independent of the specific software this group was using, it might have been anything from a simple choice list to one based on learning procedures. The listener had no way of knowing. To me the situation was inadequate; mostly because the visual and aural had no relationship by definition. Here views divide into two: those who believe that, if one were to listen with eyes closed, one should be able to ascertain the presence of a human performer, and those for whom the performer's presence is merely necessary for the production of the sound which itself may leave no clue to that presence. The example given above is thus an extreme form of the latter. So what is it beyond this primary assemblage function that makes music essentially 'live' - even in blind ('acousmatic') listening situations? ICMC 136
Page 137 ï~~Performer gesture transduction A gesture is a complex of energy profiles which we have no need to disentangle. But of course we try to. Upon the statement of one word, or perhaps at his/her own volition, a violinist alters a complex balance of parameters of amplitude, pitch, bow position (hence timbre variations and noise contribution). Are we in danger of repeating just the same set of misconceptions (not to say mistakes) as did high serialism with respect to the primary parameters of pitch, duration, and dynamic, in treating as independent, because isolateable and now measurable, different parameters of 'expression'? Let us at least try to deal sensibly with these streams of variables, and not apply rules to one type of knowledge more applicable to another. This is but an extension of the 'parameter equivalence' ideas inherited from serialism which permeate and are now perpetuated through many of the more algorithmic approaches to composition. The first generation of instrument controllers has at heart a fundamental limitation - perhaps inevitable within the commercial market. Controllers are transducers, that is they convert one form of energy (mechanical: limb and breath movement) into another (electrical signals). The process usually occurs - with the possible exception of the 'hands' developed (and now marketed) by Michel Waisvicz and Steim Studios - immediately within the controller. Measurements of variables such as breath pressure, key pressure, even timing, and for the 'hands' proximity and elevation (mercury sensors) may be made to high degrees of accuracy; but this is promptly thrown away being transduced at source to fit the limited clock and controller resolutions of the MIDI protocol. Now, this is not just another hidden attack on the limitations of the MIDI protocol per se, but a criticism that the transduction process 'physical gesture>MIDI' is not open to our scrutiny, much less our control. Translation to MIDI should be one option made at the computer, another option would be to use high resolution data for signal processing. Unless controller-produced value ranges match the range of parameter values to be controlled - and individual steps are less than or equal to the jnd for that parameter - one has a problem! Take the violinist's vibrato again, there may be several analysable components to this gesture: but beware! Are we analysing cause or effect? Human movement or instrumental result? In either case, we impoverish the analysis/transduction process and squeeze it without hesitation through standard 'seives' of pitch, amplitude (or its surrogate pressure) and discrete clocked time which, while defined by the MIDI protocol are only a mirror of our traditional Western notation system bias (cf. Wishart (1985)). This abstraction process is impoverished from the start. The pitch seive can be got around using pitchbend but why should pitch be what we wish to transduce? The absolute lack of timbral information - although some pitch trackers can output a noise component information stream - in all but the most advanced 4X type machines is evident. Thus the most we have is an attempt by rule to overcome the problem. Eg. rule: 'If a sustained sound gets louder the upper harmonics increase in amplitude'; this information is not conveyed in the transduction process, but we put it back in at the synthetic instrument. Control: who and what? Let us examine a list of freedoms and constraints for the live electronic composer and performer: Tempo and timing: Who controls the passing of time? Traditionally this has been entirely in the human domain although with increasing definition within the score tradition. While strictly 'live' electronic music did not alter this balance, the advent of 'mixed' live and pre-prepared electronic works created an extreme version of this constraint as time flow was often entirely defined external to the human performer. The problem is only partially aleviated through subdividing into 'trigger' and 'flow' components of tempo. Recent research into performance tracking relies too heavily on the 'event' approach discussed thereby effectively limiting the composer to pitch-based musics. There is nothing wrong with falling back on reestablishing a human response element within the network ie. another performer! ICMC 137
Page 138 ï~~Pre-prepared materials: We must be aware of the increasingly blurred distinction between two types of pre-prepared material: the one has traditionally been carefully created by the composer to be immutable in time (what was once called the 'tape' part) while the other is part of a developing instrumental controller synthesis tradition. The difference in musical function between the two types will not disappear simply because MIDI controllers may now trigger soundfiles from memory. Score/performance choices: Listeners are rarely aware of choice systems in composition and performance - perhaps only in memory can any notion of variability be reconstructed. Choices - even to the degree of improvisation - are there often as a motivational means for the performer as an extension of expressive variation. Timbre and expressive nuance: The need for this is well covered elsewhere in the literature (Moore 1988). To which I would add the general point that there appears to have been an overemphasis on controller technology at the expense of the extension of acoustic sources through signal processing. Summary/Conclusions In the 1980s 'live interaction' became identified to too great an extent with 'live algorithmic composition'. Material was generated largely through a crude motivic model: try to get the computer to recognise a motif and to produce a response. It really didn't matter if this response was precomposed exactly or generated through some complex algorithmic procedure, the listener had no means of knowing except vaguely in the memory of several performances. We must shift the emphasis away from this unmusical model of 'interaction' back solidly to the human performer. We have impoverished the words 'recognise' and 'determine': 'recognition' has been confined to the seived primary event parameters of Western notation (and hence the MIDI protocol), usually pitch and time, while in mirror form these have been used as inputs to 'determine' a response at the event level. We need to evolve systems which give back to the performer a sensory relation with the result, one that he/she can monitor and control through this feedback in real time - the expressive component (F. Richard Moore's 'control intimacy'). But that is not to say that composers can't write scores. Only that these and the electronic systems associated with them must not merely leave but create that possibility of performer-control at the timbral nuance and timing level. We have been seduced by the computer sequencer per se, perhaps overemphasing its role in a possible ensemble. The rush to replace human decision with computer control is misplaced. The role of computer processing on the concert platform is better used: (a) to extend the human gestures of the composer and performer and their immediate choice of and control over: sonic materials, signal processing and treatments (b) to manage, under the control of the performer, score systems in the broadest sense including systems configurations This casts the role of the computer - if you'll forgive an ironic misappropriation of metaphor - as a deus ex machina for the ensemble/performers rather than as a full member! References and Bibliography Emmerson S. (1991). "Live electronic music in Britain: three case studies." Contemporary Music Review 6 (forthcoming) Logemann, G.W. (1987). "Report on the Last STEIM Symposium on Interactive Composing in Live Electronic Music." Computer Music Journal 11(3)'44-47 Moore F.R. (1988). "The Dysfunctions of MIDI.' Coiputer Music Journal 12(I):19-28 Sialley, D. (1986). "Spectro-morphology and Structuring Processes." in Emmerson S. (Ed.) The Language of Electroacoustic Music. Basingstoke, uK: The Macmillan Press Subotnick, H. 'Ghost score' works: eg. Liquid Strata (1977), The Wild Beasts (1978), Axolotl (1981) Wishart, T. (1985). On Sonic Art. York: Imagineering Press ICMC 138