Adding Brains to Sensors

E-mail Melanie Martella

When sensor technologies are commercialized, they tend to follow a path where, in the beginning, it's all about producing the sensor element. Then, as the technology matures, this shifts to the creation of a more sophisticated and higher-functioning device that combines the sensing element with support electronics and may include some type of in-device data processing.

Over time, the ecology surrounding that sensor technology will develop. Some companies will specialize in just producing the sensing element, others will specialize in incorporating the sensing element into a packaged transducer, and yet others will specialize in bundling the transducer with additional devices to create a complete sensing system. And I'm not even mentioning the supporting industries that supply materials, testing, and calibration and other services. Pick a sensing technology and compare a modern device to one from five, ten, or fifteen years ago and you'll see what I mean.

MEMS inertial sensors have reached the fun sensor+integrated intelligence stage and it's very interesting to see how the various MEMS sensor manufacturers decide what types of processing to incorporate and how much of the processing should be in the sensor vs. handled by an external microprocessor and, if multiple sensors are integrated, how will processing needs be shared between them. This decision is heavily influenced by the desired target market for the devices. So, for example, Analog Devices has created specialized MEMS-based systems specifically for industrial vibration sensing. Freescale Semiconductor's Xtrinsic lines of MEMS sensor products, depending on whether they're intended for industrial, automotive, or consumer use, have different types of built-in algorithms and sensing capabilities. Kionix has introduced several MEMS sensor families tailored for the consumer market. InvenSense has integrated multiple inertial sensors with motion processing to create a part that synchronizes and correlates the gyro and accelerometer data and applies motion algorithms to create a sensor fusion data stream.

The truly successful applications for inertial sensors rely on capable hardware paired with good software, whether that software is in the sensor, external to the sensor, or a little bit of both. To get truly intelligent sensors that, in turn, enable a greater number of possible applications, there has to be a successful melding of hardware and software, producing good-quality data and using that data effectively.