C6.3 - Measurement Error and Statistics of Pairing Microphones for Directional Arrays in Hearing Aids

SENSOR+TEST Conferences 2011
2011-06-07 - 2011-06-09
Proceedings SENSOR 2011
C6 - Miniature Audio Transducers
D. Warren, C. King - Knowles Electronics, Itasca, Illinois (USA)
471 - 476


Directional acoustic sensing in hearing aids has been demonstrated to be an effective means of improving the understanding of speech in background noise. Directionality in hearing aids may be implemented as a pair of omnidirectional microphones, digitally processed to form a directional response. It is highly desirable in hearing aid design for the device to be small and unseen when worn, so the distance between the mircophones must be short, between approximately 8mm and 16mm. With this close spacing of microphones, the microphones should be well matched in sensitivity magnitude and phase across frequencies. The hearing aid manufacturer may choose to compensate for microphone mismatch in their processing, or purchase microphones within an extremely tight specification. This presents an expensive burden on the microphone manufacturer, as tolerances of microphone pairs for directional performance are a small fraction of the tolerances of single microphones in volume production. It is financially infeasible to reject product based on tighter tolerances, and gauge requirements for testing adequately to control the process within the tight matching tolerance is not achievalble.

The compromise is to match the microphones, selecting pairs of microphones out of production batches that have similar sensitivity magnitude and phase across a range of frequencies. The idea of measuring sufficient to match devices with a tolerance forces a different analysis of gauge repeatability and reliability than controlling a process to tolerance. The conclusion of the analysis was that the production test system needed improvement, but that improvement was achievable without burdensome ongoing cost.

Several criteria are used to determine if a pair of microphones is a match. The challenge is to maximize the number of pairs pulled from a batch. In general this is a difficult nonlinear integer optimization problem, but a simple algorithm based on graph theory comes close to the optimum solution. The number of pairs can be approximated from a product of a function of the ratio of production variation to matching tolerance for each independet variable.