Today, Monday April 17, we meet Richard Causton and Per Bloland in the spacious Studio 5 at IRCAM. They are both plunged into the entrails of an open grand piano, like a patient on an operating table. The lid has been removed, and the frame is adorned with a strange metallic paraphernalia, to which a good dozen electromagnets are attached. The two composers-in-research carefully and apprehensively carry out various experiments. Exchanging hypotheses and suggestions, they move the electromagnets, actuators and sensors longitudinally along the strings, testing different locations to achieve optimum results, sometimes with tape or by piling magazines on the soundboard.
At regular intervals, the two feel the heat sinks around the coils, as well as the capacitors and other electronic equipment, to make sure they haven't burned out due to overload and, above all, that they haven't started a fire.
Per Bloland and Richard Causton at IRCAM's studio 5
Sitting at a table not far from the large concert Steinway, in front of his computer, is Henri Boutin, a researcher in IRCAM's Sound Systems and Signals: Audio/Acoustics, InstruMents team and a specialist in the active control of musical instruments - he did his doctoral thesis on the active control of a xylophone. As the two composers whirl around the instrument, with almost childlike enthusiasm mixed with cautious care, Henri Boutin sheds light on his role in their artistic research project.
"The simplest comparison we can make of active control is that of a swing. If you push a swing against the beat, it stops. If you push it in phase with its natural frequency, you increase the amplitude of its swing. In between, there's a whole range of possibilities (depending on the force applied, its frequency and its phase in relation to the swing's oscillation) for controlling the swing."
"Active vibration control finds a wide field of application in architecture, and more specifically in the anti-seismic standards put in place in Japan, for example, to protect buildings from earthquakes. Once the structures had been placed on dampers, the question arose as to whether "anti-vibration" could be applied to cancel out the ground vibration, and ensure that the building moved as little as possible. Active control, known as "acoustic control", was originally tested in the 1970s to cancel out noise in the nozzles. Today, one of the best-known applications of active control is noise reduction, with which many active headphones are equipped. This is based on exactly the same principle: acoustic pressure (noise) is measured and the opposite acoustic pressure is injected into the headphones...".
"When it comes to active control of musical instruments, one of the pioneers was my thesis supervisor, Charles Besnainou. In the 1990s, rather than simply cancelling a vibration, he came up with the idea of amplifying or modifying it, by changing its resonance frequencies, whether on the fundamental component or the partials (thus changing the pitches of the note produced or its timbre). His first experiments were carried out on simplified models of instruments, such as a composite xylophone blade or a vibrating reed. My thesis involved designing active control algorithms to apply to a xylophone blade and a violin - using pairs of piezoelectric transducers: one as an actuator, the other to measure the resulting signal, in order to set up a feedback loop."
Henri Boutin during his thesis defense Application du contrôle actif à l'étude des instruments de musique in 2017 at Ircam
"The particular challenge of this research is that, instead of intervening on the piano soundboard, we intervene directly on the string, seen as the exciter, upstream of the soundboard. We all know how a string behaves! At least since ancient times. Its first eigenmodes are almost harmonic and highly resonant. As a result, its frequency response is made up of well-isolated peaks, enabling us to modify their position, i.e., the pitch of one partial, without modifying those of neighboring partials. Obviously, the greater the frequency variation required; the more power needs to be sent to the actuators!". The first stage of the research therefore involves measuring the eigenmodes of each string of the electronically augmented piano. A study that is complicated by the fact that, in reality, the string never corresponds to the ideal model - which, incidentally, is what makes it so rich in timbre!
"Unforeseen things always happen due to the limitations of theoretical models - but the idea is to mathematically guarantee that active control does not trigger undesirable phenomena such as feedback."
"The other major challenge of these measurements is that, to better represent and control the system, we need to think of it as a whole: i.e., the instrument AND the on-board device. Thinking of them separately might seem simpler (because we know how a string or a vibrating surface works), but it's actually incredibly easy to integrate into the system to be controlled everything that isn't modeled, such as the measured responses of the actuators, or the latency of the controller."
The system studied is no longer just the string, but the controlled string. So with the sensors and actuators, which must be placed judiciously: away from the "nodes" that correspond, for each partial, to the immobile points of the string, because then we wouldn't have access to the richness of the spectrum. We try to place them at one end, where we can act on, and measure, as many modes as possible. "On the other hand," asks Henri Boutin, "shouldn't we also take into account the strings adjacent to the activated string? And what would happen if we put several actuators on the same string?"
At the very moment Henri Boutin asks the question, as if to illustrate the conversation, and yet without having followed it, Per Bloland and Richard Causton try the experiment: they move the activators longitudinally along the string, even try to place them below the string, try to activate three sensors on the same string (an F sharp), switching them on and off one after the other. Strange sounds emerge from the piano, which seems to have a cat in its throat.
Set up on a grand piano
As Henri Boutin mentioned earlier, another problem is latency. To control a vibration properly, the signal sent to control it must be exactly in phase with it - otherwise, if it's out of phase, the control won't have the desired effect. However, between the moment when the sensor detects the vibration and the moment when the actuator acts, the signal must be transported and analyzed, the computer must calculate its action, which in turn must be controlled, and its signal transported to the actuator, whose action is not immediate either. Not to mention any buffering of the signal by the computer.
"This is one of the main challenges of control," says Henri Boutin. In practice, we make use of the fact that vibration is periodic: rather than being exactly on time, we delay the action signal so that it synchronizes with the next period. But this also presupposes perfect knowledge of this period, and the right sampling to encode the signal in order to achieve this: theory tells us that this sampling must be greater than 10 times the signal's maximum frequency. For our purposes, the audio standard of 44100 Hz is sufficient. This latency issue also means that the control of attack transients is the most complex to achieve. This problem has been tackled in the past by several members of the Sound Systems and Signals: Audio/Acoustics, InstruMents team, who have developed technology with the aim of minimizing this latency."
Performance of Of Dust and Sand by Per Bloland
If all goes well, the pianist may well play an A, which will sound like an A sharp - or any other note, for that matter, and even a melody. Henri Boutin explains: "While playing, the system's modes can be changed, the aim of the game being to develop an interface enabling the performer to control them - a device which, moreover, must be able to be transported and easily adapted to any instrument, at the whim of a production, after calibration." He concludes that, "in theory, anything is possible - you could produce any harmonic spectrum on any string, whenever you want - but on the other hand, the device may have certain limitations."
In a forthcoming episode, we'll meet Brigitte d'Andréa-Novel, director of the laboratory.