Final Activity Report Summary - NEST-PM (Nonlinear effects in string instruments: perception and modelling)
The main goals of this research were to assess the perceptual importance of nonlinear components in string instrument sounds and to develop more elaborated modelling techniques. The nonlinearly generated longitudinal string vibrations are particularly important in the case of the piano, where they are believed to be responsible for the metallic character of the tone for low keys. Formal listening tests were conducted showing that the longitudinal components are perceivable for keys below the middle register. The results can be directly used in sound synthesis of piano tones, as they give an indication for which notes the phenomenon should be modelled.
Additionally, the mechanical admittance properties of guitar and piano bridges were measured; giving an indication to what degree the different polarisations of the string vibration are coupled. The sound pressure radiated by the piano soundboard and the guitar body was also measured by exciting the bridge in the directions corresponding to the three polarisations. These results can be used for the parameterisation of physics-based sound synthesisers. The bridge of an instrument is a passive system, meaning that it can only dissipate energy, and cannot generate it. However, when the measurement data of a bridge is used for the parameterisation of a sound synthesis model, the resulting instrument model often corresponds to an unstable system, due to inaccuracies in measurement and modelling. For ensuring the stability of the sound synthesis model, a novel admittance model was developed, that produces inherently passive impedance models even if there are some accuracy problems in the measurements. In addition to the direct applicability to musical acoustics and sound synthesis, it is foreseen that the passive admittance model can be used in other fields of acoustics, too.
The passive admittance model is based on a new design algorithm for the parallel second-order filter, where the poles are predetermined and only the zeros are left free for optimisation. Besides admittance modelling, the parallel filter has been used for the efficient and flexible modelling of the body of musical instruments, which can be applied in real-time sound synthesisers. The efficiency is originating from the fact that the frequency resolution of the filter can be adjusted freely by varying the poles of the filter. Here a perceptually motivated approach was taken, that is, the frequency resolution was set according to the frequency resolution of the hearing. The parallel filter has also been applied in equalising the frequency response of loudspeakers. In both applications, the parallel filter requires less additions and multiplications compared to the previous method of same performance. Since the parallel filter has been found to be useful in many applications, it can be considered as the most significant scientific achievement made over the course of the project.
Additionally, the mechanical admittance properties of guitar and piano bridges were measured; giving an indication to what degree the different polarisations of the string vibration are coupled. The sound pressure radiated by the piano soundboard and the guitar body was also measured by exciting the bridge in the directions corresponding to the three polarisations. These results can be used for the parameterisation of physics-based sound synthesisers. The bridge of an instrument is a passive system, meaning that it can only dissipate energy, and cannot generate it. However, when the measurement data of a bridge is used for the parameterisation of a sound synthesis model, the resulting instrument model often corresponds to an unstable system, due to inaccuracies in measurement and modelling. For ensuring the stability of the sound synthesis model, a novel admittance model was developed, that produces inherently passive impedance models even if there are some accuracy problems in the measurements. In addition to the direct applicability to musical acoustics and sound synthesis, it is foreseen that the passive admittance model can be used in other fields of acoustics, too.
The passive admittance model is based on a new design algorithm for the parallel second-order filter, where the poles are predetermined and only the zeros are left free for optimisation. Besides admittance modelling, the parallel filter has been used for the efficient and flexible modelling of the body of musical instruments, which can be applied in real-time sound synthesisers. The efficiency is originating from the fact that the frequency resolution of the filter can be adjusted freely by varying the poles of the filter. Here a perceptually motivated approach was taken, that is, the frequency resolution was set according to the frequency resolution of the hearing. The parallel filter has also been applied in equalising the frequency response of loudspeakers. In both applications, the parallel filter requires less additions and multiplications compared to the previous method of same performance. Since the parallel filter has been found to be useful in many applications, it can be considered as the most significant scientific achievement made over the course of the project.