In audio processing, such as data compression and time/pitch scale modification, it is necessary to model the audio signal using an appropriate 'analysis' model. The 'sines+transient+noise' model, where the signal is separated into components of harmonic, transient and residual noise, is such a model that can be used in the analysis. In this model the harmonic separation is achieved by sending the signal through a uniform set of filters, usually having a fixed frequency resolution (bandwidth). In this paper we propose a pitch-synchronous IIR Laguerre filter having a time varying bandwidth (constant Q-factor) for the filter set. The advantages of using such a filter set in audio processing are discussed in the paper.