The following article was originally published in Computer Music Journal 2(3):24-29, 1978; reprinted in the book Foundations of Computer Music (p. 588-600, MIT Press 1985).
"Music for an Interactive Network of Microcomputers" John Bischoff, Rich Gold, and Jim Horton
We describe here the music presented in concert on July 3, 1978, at the Blind Lemon, a music gallery in Berkeley, California.
An Overview of the Piece
Traditionally, music has involved more than one person, either in its composition, in its production, or in both. In fact, it seems to be one of the most social art forms. Although there has been individually produced music as well, computer music by its nature could until recently only be individual, solitary music. However, with the introduction of microprocessors at a reasonable cost, composers can now own microcomputers, and true computer bands, free from major institutions, are possible. Though such bands can take many forms, network music seems the most suitable and contemporary.
All three of us owned KIM-1 microcomputers, but aside from the fact that they simplified many of the input/output problems they were not significantly similar. Each composer had programmed his computer with a music program that was by itself able to produce music; however, the programs were also able to input data that would affect the musical content and to output data that would affect another computer's program. Each computer had its own music output, either to a digital-to-analog converter (DAC) or to digitally controlled electronics.
It was decided that for the first concert a simple formation would be used. In this case, each computer sent data to one other computer and received data from one other computer, so that a circular data structure was effected. How the received data were used and what data would be sent were the individual composer's choice, though the bus structures were mutually agreed to by each pair of composers. The final musical output was mixed together and broadcast over a high-fidelity music system.
The exact configuration used during the concert was the following: Bischoff sent data to Horton, Horton sent data to Gold, and Gold sent data to Bischoff.
Figure 33.1 - As can be seen here, the basic flow of information was circular. Bischoff's KIM sent one line of serial information (which served as both interrupt and data line) to Horton's KIM. Horton placed four bits of parallel information onto a latched data line for the use of Gold's KIM, while Gold gave information to Bischoff, also on a four-bit latched data line. Each of the three KIMs had its own musical output: Horton's machine featured digitally controlled circuits; Gold's and Bischoff's outputs were both direct digital-to- analog conversion. The three musical outputs were summed together and played to the audience at a goodly volume.
The Individual Programs
John Bischoff's program was originally composed for a performance with Phil Harmonic at the WORKS Gallery in San Jose, California, in February 1978. The performance environment was casual, attentive, and allowing of social interaction, and the music was designed around the idea of long moments of rest interspersed with computer tones. The occasional tones generated by the KIM served to punctuate the performance. The periods of rest between tones lasted up to one minute, and, as Phil Harmonic pointed out, one could even forget that the KIM was running. As heard in the network, the program played somewhat the same complementary role but with more emphasis on the ensemble music properties of the three parts moving in relation to one another.
Figure 33.2 - The musical flow of Bischoff's program CUE HORTON - PLAY THE SOUND - PICK A LONG SILENCE - PICK A TIMBRE - PICK A PITCH - WILL TONE SLIDE? - PICK TONE DURATION - IS SILENCE OVER? - NO, KEEP WAITING; YES, GO TO BEGINNING
All the choices in figure 33.2 are based on a continually renewed string of random numbers. The output waveforms are routed through and 8-bit DAC and are of a constant amplitude. There is no predetermined sequence of pitches, as each run-through of the program involves on rest period followed by one pitch event. There are four possible waveforms: sawtooth, triangle, and two types of random waveshapes. The random waveshapes generate particularly striking timbres and noticeable sidebands during pitch slides. (Since 1978 this program has been modified and developed extensively. It is currently run simultaneously on two microcomputers, and is dynamically performed as a piece called "Audio Wave").
In April 1978, Horton modified his program so as to accept data regarding the specific frequencies that Bischoff's computer was putting out. Bischoff altered his program to enable it to send these data each time it was ready to produce a tone.
A single line was connected between the two computers to act as both an interrupt line and a serial data line. Before each tone and before each rest, and interrupt was sent. This was followed by one bit of data, which acted as a flag to indicate the upcoming event. In the case of a tone, a floating-point representation of the frequency was transmitted, serially, following the flag bit. On receiving that, Horton's computer calculated and played pitches that were in justly intoned harmonic intervals to the note it had received. His program jumped back into its independent mode when it received an interrupt followed by a rest flag.
Bischoff's computer was also connected, by four parallel lines, to Gold's computer, the information from which was used to influence frequency or rest duration, or both, or was ignored.
From the equation
where x is the latitude and y the longitude of a traveler on a fairly smooth, basically continuous surface, where z is the altitude of that traveler, and where the traveler exhibits a continuous, closed motion about the surface, z can be shown to exhibit periodic-wave-like properties where the frequency of the wave (z subscript w) is determined by the length of the traveler's closed walk and the speed of the walking while the timbre of the wave, including amplitude, is determined by the hills and valleys of the land. The problem would become far more complicated if the traveler could move at more than one speed; however, since in Gold's program the traveler can move at only one constant speed, frequency is determined by the length of the walk only. Further, if that traveler moves at a speed such that the periodicity of z subscript w usually falls between 20 and 20,000 Hz, and given the appropriate transducers, there is music.
Note to Figure 33.3 - A block diagram of Gold's "Terrain Reader". Each of the four upcounters controls one of the four endpoints of the two up-down counters. The upcounters are updated at the beginning of each new note. The up-down counters count at audio rates and specify a point of the surface or terrain, which is held in one page of memory. The value of the point specified is sent to the DAC. The tune, or the series, of notes that results, is determined by the eight endpoints of the upcounters, which are set, in this version of the piece, by the information flowing from Horton's computer.
The entire program, save a latched output port and a simple DAC, was contained within the KIM's 1 Kbytes of memory. The surface f was modeled in a 16x16 matrix, occupying page 3 of memory, and behaved like a land on a torus. The continuous closed motion of the traveler was produced in a Lissajous-like fashion using two software up-down counters (triangular outputs). While the rate of update of the two counters remained constant, the four endpoints (a top and a bottom for each counter) were under program control. The fundamental frequency of z subscript w is determined not only by the length of the count, but also by the relationship between the lengths of the two up-down counters. Further, since the pattern could be relocated anywhere on the surface f, there was a way of altering the timbre of a given pitch.
The endpoints of the up-down counters were determined by four up- counters, one controlling each of the four endpoints. These were updated at the beginning of each new note or sound. The endpoints of these four upcounters were controlled by data input from Horton's machine. The "tune" was determined by the series of endpoints generated by the up-counters, given that at the beginning of each note they were all incremented by 1 and wrapped around (i.e., rotated) upon reaching maxima. The length of each note, the number of notes in each "song", and the length of silence between songs was determined from information coming from the surface. The durational information was sent to Bischoff's machine (as a four-bit word).
The program, called "Terrain Reader", was part of a broad piece entitled "Fictional Travels in a Mythical Land". What the "Terrain Reader" reads is the land, the shape of which was determined by the general "myth" from which the entire "Fictional Travels" piece was derived. That is, the program was not intended to be a general- purpose music program but rather to be an integral part of the piece itself.
Horton's program did two separate but related things, one harmonic and the other melodic. the melodic program was written out of curiosity in order to listen to an aspect of Max Meyer's psychological theory of melody (Meyer 1901). Meyer's empirical investigations led him to conclude that no tone is in a specifically melodic relation with another unless the interval between them can be represented by one of the rations 2-2, 2-3, 2-5, 3-5, 2-7, 3-7, 2-9, 2-15, 5-7, and 5-9, or else they are both related to a third tone by one (not necessarily the same) of these ratios. Meyer's notation represents classes of ratios (for instance, 2-3 indicates 3/2, 4/3, 3/1, 8/3, etc.), because according to his observations octave transposition does not make any difference in the kind of relationship perceived.
Meyer defines the "complete musical scale" as "the series of all tones which may occur in our melody, however complex this may be." He shows that according to his theory it "is represented by the infinite series of all products of the powers of 2, 3, 5, and 7." However, in his extensive analysis of existing melodies, including those of the highly chromatic music of his contemporaries, he found that 29 tones suffice for a complete description. None of these tones has a factor of 2 > 2 to the 10th, of 3 > 3 to the 6th, of 5 > 5 to the 3rd, or of 7 > 7 to the 1st. This 29-tone-to-the-octave scale is arranged in figure 33.4 so that, with a little perusal, the melodic relationships are evident.
Note to Figure 33.4 - Meyer's empirically derived, 29-tone-per- octave scale arranged by fifths vertically, by major thirds horizontally, and by septimal minor sevenths perpendicular to the page. In order to project these tones into an octave they should be multiplied by some power of 2 so that they fall within the range 512-1,024. Each pair of tones can be seen to form two intervals with each other. Only a subset of pairs falls within the class of the ten melodic intervals.
The program works by randomly selecting a note from the scale and calculating whether or not it and the note already sounding form an interval that falls within one of the ten melodic classes. Phil Harmonic has commented that this system "usually seems right on the edge of breaking into a recognizable tune."
Range, tempo, "rhythmic pattern," and density of rests are determined by simple algorithms whose parameters can be changed while the program is running. This is done by switching subroutines in or out or by changing values in memory. The program is designed so that any constant or parameter that might conceivably be changed is located in page 0 (0-FF hex). A section of the program allows the player to use a hex keyboard and an LED display to inspect any location in page 0 and to enter new data into a buffer. The player can then transfer the contents of the buffer to a memory location at the right musical moment.
Whenever a new note is played, data about the current "rhythmic pattern" is latched into an output port for the use of Gold's program. Audio is obtained by using an LSI device, the 8253 programmable interval timer. This chip contains three 16-bit down counters and a control-word register. Each counter is configured as a square-wave rate generator whose frequency is set by dividing a 1-MHz clock by a number supplied to it by the computer.
Any counter is always at the same pitch as the others but at a slightly different frequency. One is played at the scale frequency, another is offset from the first by a small fixed amount, and the third is offset by a randomly determined amount. The three components of the sound are mixed together to produce precisely controlled flanging.
(The following section was written using a technique very similar to the process used by the three computers in the "Network Piece" discussed in this article.)
The event of three composer making music together using ideas and structures developed independently without thought of future collaboration now seems a natural musical process. This is due in large part to the work of John Cage.
Because nobody is only an ear, the sound of music, bracketed apart from the projection of socially relevant images and meanings, is, while often quite interesting, not necessarily the main focus of a composer's work.
Very high technology is about working together in large-scale teams, e.g. the space program. It should be no different for modern music.
Independent simultaneous activities viewed as one single activity always bring to mind the idea that groups can work wonderfully together without the anxiety of control structures that supposedly ensure success.
Yes! What a pleasure to play and be part of a dynamic musical cybernetic process! To explore catastrophe hypersurfaces in the relative safety and comfort of involvement with one's friends and neighbors!
At each stage in the development of the network the music changed unpredictably. It became clear that it was impossible to tell beforehand where the music was going to come from.
At this stage in the development of the experimental music tradition it is thought well to develop a personal, even idiosyncratic, approach to music. To find such an approach is not always easy. The advent of not-very-expensive microsystems can help free the computer musician from the pressure to conform to the mores of highly structured business and academic institutions.
It seems obvious that three composers would write different music for subroutines in an IBM mainframe computer than for micros they personally own in a network. I suppose that if we hooked these routines together and the result was a Bach fugue with perfectly synthesized strings, we'd have to rewrite the programs.
Although the network seemed to have a sound more characteristic of one active musical intelligence, it could be viewed as three people making music and listening to each other continually along the way.
To bring into play the full bandwidth of communication there seems to be no substitute, for mammals at least, than the playing of music live.
For music exhibits the properties of both gyroscope and steering rocket for a society.
For instance, having one's own microcomputer reduces the need for contact with institutions just to do one's music, while at the same time it encourages collaborative work between artists. This latter situation is created by the possibility of everyone on the block owning roughly the same device.
It was John Cage who pioneered an important form of collaborative music: the simultaneous playing of compositions. An extension of that idea is to write "reactive" compositions that can interact with one another as well as with their players. This approach makes possible a collective style of music while allowing each composer the opportunity to invent and play complete designs not necessarily subordinated to other parts or wholes.
There are many ways of handling form other than putting the largest and fewest structures at the top and the smallest and most numerous at the bottom. In this case forms were distributed fairly evenly throughout.
Computers seem to start from such a low level of musical intelligence (in contrast with the impression that synthesizers give immediately) that the potential of modeling musical intelligence using computers appears promising.
However, at present, the philosophies guiding the development of general-purpose software systems and programs can be questioned. For instance, why the great effort to synthesize the sound of the violin and the piano? Why not the koto, the accordion, the Peyote ceremony's rattle? For that matter, how many composers are really committed to the idea that art should imitate nature anyway?
Music has that wonderful ability that when you have three pieces of music working together you still have music.
Though synthesizers always offered the potential of multisynthesizer group music, and there are some nice examples, microcomputers seem to fit the group-music situation even better. One can show up at a rehearsal or a performance with little more than computer in hand. Microcomputers are conceptually both a module and an entire system.
And ideally music should contain within itself all the information most important to a culture. An orchestra especially should have a structure that exhibits the best types of sociopolitical arrangements imaginable.
Though a single computer, micro or macro, is regal in nature, with its hierarchy of registers, a network of them isn't necessarily.
(Live concerts always seem to have a shared feeling between the performers and audience that makes all the startings and stoppings enchanting.)
"The patterns of control in a system tend to reproduce the organizational chart of the institution that designed the system." Hierarchical design derives from the myth that militaristic kings are better at getting things done. The theory of heterarchical and anarchistic systems design has been underexplored, although experimental musicians have been engaged in the processes for years.
John Cage set up a new minimum, on the potential plane of music, away from the classical music valley but close enough to draw composers away from it. This allowed for the direct modeling of contemporary ideas (if we want to talk about networks, we build a network) and the use of the available technology for sound production.
The inexpensive microcomputer is a decentralizing influence on the way art involving technology is structured in society. This might help to balance the inherently centralist tendency of the arts and music in general.
Since computers are now as portable as other musical instruments, it is easier to think about grouping them into bands and orchestras.
In our band, each computer contained the program of one composer and produced a sound of its own. Furthermore, all three computers played those programs simultaneously and in the same real time. Beyond that, they were interactive in that each affected the next. That is to say, the piece unabashedly manifested all major trends in music composition of the last 5,000 years.
When the elements of the network are not connected the music sounds like three completely independent processes, but when they are interconnected the music seems to present a "mindlike" aspect. Why this is so or why we can perceive some but not all activities as the product of an artificial intelligence is not understood.
The nonhierarchical structure of the network encourages multiplicity of viewpoints and allows the separate parts in the system to function in a variety of musical modes. This means that the moment-to- moment form the music takes is the result of the overlapping individual activities of all the parts, with the coordinating influence of the data exchanged between computers.
Since mistaken concepts and "bugs" seem inevitable, and since plans of any complexity usually break down, it is heartening to note the mystery that where several errors intersect they very often make an interesting pattern.
But how do you get three modern composers to work together? Micros, with their simple structures, provide an answer. On the other hand, large computer facilities and large electronic music studios seemed to be an extension of the older romantic idea of the individual composer writing notes in isolation from an audience and other musicians.
The structure of a circular system satisfies the desire for a symmetrical interactive network where the flow of influence emanates evenly from each point in the system.
Because musicians, since time immemorial, have been playing together, music has developed into a wide variety of "naturally occurring" parallel processing systems.
We created an interesting creature and spent an evening, in public, listening to it.
Editor's Note: A recording of the League of Automatic Music Composers is available from Lovely Music/Vital Records, New York: "Lovely Little Records" (VR-101-06) music by John Bischoff, Paul DeMarinis, Phil Harmonic, Frankie Mann, Maggi Payne, and "Blue" Gene Tyranny.
Meyer, Max. 1901. "Contributions to a Psychological Theory of Music", Volume I. University of Missouri.
- typed by John Bischoff 11/96