Characterization of a MEMS Directional Sound Sensor

Naval Research Laboratory

Sunday, February 01 2009

Operation of the sensor is based on the hearing organ of a fly.

There is a wide range of potential military applications in which ambiguity in bearing occurs with respect to sound. For example, autonomous unmanned aerial vehicles (UAVs) could employ a sensor to determine the bearing of an explosion and conduct battle damage assessment (BDA) on it. With existing sensors this is difficult to do because the explosion is too short in duration to use the Doppler effect to determine the bearing. Also, an autonomous underwater vehicle (AUV) acting as a quiet platform to tow a short, omni-directional hydrophone array must contend with bearing ambiguity.

A directional microphone has been developed that eliminates the bearing ambiguity and was added to the towed array. The integrated system constitutes a valuable acoustic measurement tool.

Characterization of a directional sound sensor has been determined using micro-electromechanical systems (MEMS) technology based on the hearing system of a small fly (Ormia ochracea). The fly uses coupled bars hinged at the center to achieve the directional sound sensing by discriminating the vibration amplitude of each bar. The sensors used in this case were fabricated using Silicon on Insulator Multi-User MEMS Processes (SOIMUMPs) technology available through MEMSCAP.

The wings of the sensor were made primarily of solid silicon plate in order to maximize the vibration amplitude when the substrate underneath them was removed. A set of sensors with perforated wings was also included in the design to gauge the sound coupling efficiency. The sound sensor was found to have two resonant vibrational modes (rocking and bending). The sensor was simulated using finite element analysis and tested by actuating the sensor using a sound stimulus. The purpose of the simulation was to obtain a valid representation of the sensor under study that could be used in future sensor development.

An analysis to describe the relationship between the sensor's amplitude of vibration and various parameters as the angle of incidence and the intensity of sound was conducted. Experiments, as well as simulation using finite element software, were conducted to assess the performance when two prototypes located on a single chip are tested under varying conditions.

The experimentally observed vibrational frequencies were found to be in good agreement with those of the simulated sensor. The amplitudes of vibration were found to be of the same order of magnitude compared with the simulated sensor and significantly larger than values reported in previous studies that employed sensors fabricated using the PolyMUMPs process. The amplitude of vibration was found to increase as the incident angle was increased and followed in good agreement the theoretical predictions.

Some differences between the two prototypes were found, especially as the frequency diverges from the resonant frequency (2,980 Hz) of the sensor. This analysis points out some disadvantages of the current setup of the physical experiment. Some changes regarding the position of the sensor and the absorbing material that was used were made to attain more reliable experimental units. The model developed in this work used as a response variable the natural logarithm of the vibration amplitude of the sensor. In order to find a goodness-of-fit measure that applies to the response variable directly, estimates on the logarithmic scale were converted back to the original units.

Because a logarithmic transformation was used, it is appropriate to consider a measure that expresses the explanatory power of the model in relative terms. The statistical model developed achieved an average relative error (ARE) of 3.80 percent, which implies that the model was capable of predicting the vibrational amplitude of the sensor with an error that averages 3.80 percent of the actual value. This suggests that the model provided an adequate representation of the behavior of the sensor.