Chatter detection has assumed a central role in monitoring of machining operations. Defining optimal detection algorithm is still an open task, besides performances of these algorithms strongly rely on employed sensors: microphone sensors are gathering growing interest in this scenario, mainly due to low cost and ease of integration into machine tool environment. Nevertheless microphone response could be altered by reflection, reverberation and absorption inside the machine tool chamber, returning misleading measurements. This work is focused on the identification and compensation of microphone response, to maximize linear bandwidth, that represents an absolute requirement in chatter detection approaches, especially if based on frequency domain analysis. Identification and compensation methodologies are presented and tested in experimental tests.