Energy-balanced models for acoustic and audio systems: a port-Hamiltonian approach