The limitation of size on an electrodynamic loudspeaker’s low-frequency reproduction is an old, well-known problem. It is particularly severe in mobile telephones for hands-free speech reproduction. Many authors have described how various analogue and digital electronic techniques can extend the lower limit of a loudspeaker’s frequency range without increasing its size. One well-known technique is equalisation, in which frequencies below the loudspeaker’s lower limit are emphasised. A problem with equalisation is that the properties of a loudspeaker vary significantly due to manufacturing tolerances, temperature fluctuations, and ageing. One technique to track these changes is to perform system identification on the loudspeaker’s electrical impedance. In this project it has been found that parameters of a second-order adaptive IIR filter plant model of the electrical impedance can be used to update parameters of a digital filter performing equalisation. This was found to work for a loudspeaker in a closed-box enclosure. A fourth-order IIR plant model filter has been shown to be appropriate for identification of the parameters of a loudspeaker in a bass-reflex enclosure. In the search for efficient system identification algorithms, the behaviour of loudspeakers has been studied very closely. During a study of the single-suspension loudspeakers used in mobile phones, ‘rocking modes’ or ‘wobble behaviour’ was found to be an obstacle to using simple system identification algorithms. As a result, techniques for analysing the problem of ‘rocking modes’, along with strategies for avoiding them, were developed. The method used experimental modal analysis on data collected from a scanning laser vibrometer. A method for transforming the results of an experimental modal analysis on such data into the parameters of a rigid-body model with translation and rotational degrees of freedom is under development.
|Effective start/end date
|01/06/1999 → …
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.