[Dear friends: This page was last updated in 2003.  Please see my current research page for more recent information.]

For the past several years the primary area of my computer music research has been invesigation into methods of synthesizing dynamic musical textures with the aid of computer technology. My work has examined stochastic procedures, realtime control interfaces, and artificial life most closely. I have both investigated pre-existing computer music methods and software and have designed and written my own. The four examples of my work shown here are Appliance, Juicer, Ripples and Treembre. If you wish, you may download a PDF document summarizing these projects by clicking here. I have also written short essays on the process of composition of several of my pieces, including Gilgamesh and Reality House.

I have also completed research into the effects of user interface design on computer music composition, and techniques for building realtime computer music performance instruments. Email me for more information regarding this work.

1. Appliance: An Interactive Installation Performance

Figure 1: Performance of 'Appliance'

In the spring of 2000 violinist Maja Cerar, sculptor Thomas Charveriat, and I embarked upon a collaborative project entitled `Appliance,` which was an interactive performance installation for violin, electroacoustic music, and mechanized sculptures. In performance, the violinist wore a sensor glove to communicate with a Max/MSP patch that controlled a delay line and playback of samples. Meanwhile, she also operated a footswitch that controlled the activities of the sculptures, which were connected into a network (Figure 1). The sensor glove is worn on the violinist's right (bowing) hand (Figure 2). The glove is handmade from clingy lycra with open fingertips to allow for better bow control. Three force-sensing resistor (FSRs) are attached to the index, third, and fifth fingers of the glove, and an accelerometer (tracking X and Y velocities) is positioned on the back of the hand. Wires from the glove's sensors and the accelerometer unite at a serial connection which sends the data to a Pic Microcontroller residing in a small switchbox. The chip translates data coming from the glove into MIDI control signals and sends this out to a MIDI interface and the central computer.

The MIDI cable feeds into a Macintosh laptop running Max/MSP and a patch built specifically for this piece (Figure 3.) The motions of each finger are mapped to control an individual aspect of performance: delays on/off, samples on/off, and changing the currently active sample bank. The accelerometer data, originally intended to be used to control panning and amplitude, was eventually ignored in performance. The delay time can range from 0.125 second to 5.0 seconds and is controllable from the computer, which is operated by a second performer. The amount of delay feedback is also controlable at the computer. The samples are brief (0.25 to 3.0 seconds) and contain a mixture of violin sounds and other percussive sounds. In addition to the controller glove, a small microphone is mounted on the violin and this signal is also fed into MSP. Using the Miller Puckette’s fiddle~ object, we track note onsets, envelopes, and pitch. This data is used to trigger samples when the sampling function is turned on (fifth finger's FSR.) We use the MSP groove~ object for sample playback, looping short percussive samples to match violin notes' amplitude envelopes and transposing samples according to the estimation of the violin's current pitch.

Fig. 2: The sensor glove

Each sculpture is composed of motorized elements mounted on a sheet of metal and elegantly laid into a burnished aluminum suitcase. Most of the mechanisms of the sculptures create both motion and sound. Each sculpture also contains at least one microcontroller chip, programmed in BASIC to define the behavior patterns of that sculpture. Because of limitations of both the mechanics of the sculptures and the memory of the microcontrollers, each sculpture’s behavior is limited to being a set of loops of rhythmic patterns.

Because the microcontrollers were not designed with musical applications in mind, programming anything more complicated than a metronomic beat becomes quite time intensive. In order for the microcontroller to play a rhythmic patter one must calculate the exact number of milliseconds between each subsequent note onset in the patter, type in BASIC code using these values for every note of the music, and then load this file into the microcontroller. Obviously, this method of composing musical ideas is quite non-intuitive and can be both frustrating and time-consuming, as the composer cannot hear what s/he has programmed until after the code has been loaded into the microcontroller. To make the process easier, we created interfaces reminiscent of an analog sequencer in Max/MSP to program the sculptures. These allow one to quickly compose rhythms with both auditory and visual feedback. This intuitive interface makes it easy to set up sophisticated polyrhythmic patterns, and then at the push of a button Max writes the BASIC code needed to program the microcontroller.

Figure 3: Main 'Appliance' Max/MSP patch

The sculptures are connected into a network to coordinate their activities. The violinist operates a MIDI footswitch to cycle though a preprogrammed pattern of behaviors of the sculptures. When activated, the mechanism of a sculpture begin to move and make noise. Each sculpture is miked, and this signal is mixed with the violin and the MSP output for amplification.

Each performance of the piece is highly improvisatory. The intention was to set up a system of possibilities for the musician to explore-there is no notated score at all. We made two public performances of the piece, and the second of these was recorded onto video (available upon request.) We have placed more documentation on this project at the website of sculptor Thomas Charveriat. 

---

Other examples of my research include Juicer, a realtime DSP instrument; Ripples, a stochastic composition system; and Treembre, a SGI/Linux application. The Treembre, an attempt to hierarchically organize elements of timbre and animate them in time.