Doctor of Philosophy
Faculty of Creative Arts
Havryliv, Mark, Haptic-rendered practice carillon clavier, Doctor of Philosophy thesis, Faculty of Creative Arts, University of Wollongong, 2012. https://ro.uow.edu.au/theses/3719
The carillon is one of the few musical instruments that elicits sophisticated haptic interaction from experienced and inexperienced players alike. The fact that practice instruments do not reflect the idiosyncratic force-feedback of individual carillons creates distinct problems for both types of player. The light touch, consistent across the range of a typical rehearsal instrument, limits the rate at which inexperienced players can develop physical stamina and musical intuition for the force-feedback and related timing constraints of the real instrument; rehearsal instruments are currently the only way novice players can learn, create, or practise musical arrangements in private. Experienced players, less reticent about rehearsing in public, encounter a variant of the same problem: preparing for concert performance on an unfamiliar instrument with little opportunity to adapt to its idiosyncrasies.
The development of an electro-mechanical haptic carillon baton in this thesis is structured to address these issues. A multi-body dynamic model of the carillon mechanism is developed based on measurements and analysis of the mechanism for bell 4 at the National Carillon, Canberra. This model is extended to the rest of the instrument by accounting for variation in physical parameters in clappers across the range of the instrument; a linearisation of this model is also derived and validated. Variation in physical parameters in clappers is demonstrated to correlate strongly with the relationships identified in the existing literature on the design of carillon bells; this further generalises the instrument-wide model for variation in physical parameters, allowing for a priori estimates of clapper dynamics based on an instrument’s range alone. This instrument-wide model is combined with dynamic and static measurements that encapsulate the motion and force-feedback characteristics for individual batons across the range of the instrument; taken together, these form a carillon’s haptic signature. These characteristics of the carillon are susceptible to environmental factors and mechanical defects. In cases where the dynamic model fails to account for all elements of the haptic signature, for any individual baton these elements can be modelled by a novel implementation of the Discrete Wavelet Transform.
This extended virtual model is validated by comparing its predictions with experimental data. Forward dynamics simulations across the entire range of the instrument demonstrate that the model replicates clapper and baton motion in response to a step force input, and when the baton is fully depressed then released. Additional inverse dynamics simulations compare favourably with position-force data.
A haptic device is developed in order to validate the model against carillonneur perception. The device is designed to be retrofitted to existing rehearsal instruments and a novel method of sensorless force sensing is developed. This method determines user-applied force by analysing the current command signal output from a commercial position-control servo to the linear actuator. The noisy signal is filtered using a user-tuneable Kalman filter built on a state-space model of the servo and actuator derived from a system identification procedure applied to the servo system. This filtering system rejects high-frequency noise associated with the operation of the actuator itself, but remains highly-responsive to sudden user-applied gestures.
The generalised model and the performance of the haptic device is evaluated by carillonneurs from the National Carillon. A quantitative evaluation is conducted which requires carillonneurs to estimate which of the 55 batons is being simulated by the device; results from this evaluation demonstrate that the model and device successfully replicates varying force-feedback across the instrument’s range. Qualitative feedback indicates that the dynamic model accurately simulates the feel of individual batons.