Thermal Modelling of Voicecoils in Microspeakers

Abstract: Microspeakers can overheat and break if not monitored and regulated. This monitoringis usually done by adding a pilot tone that introduces energy to the signal.A problem with this approach is the slow update rate of the temperature estimate.This in combination with a fast temperature rise could result in an audible regulationof the input. By simulating the voice coil temperature these problems couldbe mitigated. In this thesis, two existing grey box models and one novel black boxmodel are estimated for different speakers and evaluated using different signals.The results are promising and indicate that all models can estimate the voice coiltemperature with a mean error below one degree. The tests show that a correctinitialization of the model is crucial. Therefore the suggestion to Cirrus Logic,who hosted this thesis project, is to combine a feedforward model with eithertemperature sensor data from the mobile device or a pilot tone.