Integrated smart systems with MEMS sensors
STMicroelectronics, Coppell, TX
Each MEMS sensor consists of three parts: a MEMS sensing structure, an application-specific integrated circuit (ASIC) and the device package. The sensing structure part is responsible for detecting the capacitance or resistance change when the proof mass moves from the center position due to the external motion or force applied. The second part, which is an ASIC, consists of a charge amplifier to convert the output of the mechanical sensing part into an analog output voltage that can be digitized through an A/D convertor and available to the user in the digital format. The package of the device, in addition to providing the housing to the sensing and processing die, also plays an active role in the performance of the device contributing to its stability over temperature and time.
The host processors in smart systems (e.g., smart phones and tablets) have limited resources for sensor data acquisition and processing. Therefore, MEMS sensors need to include more computing power and embedded features to reduce the load of the host processors.
Most MEMS sensors have a built-in self-test feature. The self-test can be used to verify if the sensor is functioning after PCB assembly. This functional test doesn't require physically tilting or rotating the PCB for inertial sensors.
|Inside a MEMS accelerometer and gyroscope|
Most MEMS sensors have one or two interrupt output pins available for connecting to the GPIO ports of the host processor. The host processor is not required to keep acquiring the sensor data to determine the current status of the device. Instead, the sensor is running in the background. When the predefined criteria are met, the sensor will generate an interrupt signal on its output pin to notify the host processor. The host processor can then decide if this interrupt needs to be serviced or not.
First-in-first-out (FIFO) is another feature for power saving which can be implemented in ASIC. The host processor doesn't need to acquire sensor data all the time. Instead, the sensor can collect data and store it into the FIFO in the background.
When the FIFO interrupts are generated, the host processor can wake up and read all FIFO data samples at once. Then the host processor can process the sensor data to see if further action needs to be taken.
As some interrupt features embedded in an accelerometer cannot distinguish fake motion from the real one, the processor needs to acquire sensor data to determine the nature of the motion. Future smart sensors will have more advanced computing power such as finite state machine (FSM) for reliable interrupt generation.
A low power microcontroller can be integrated into an inertial module unit to run the sensor fusion algorithms so that the final dynamic accurate pitch
oll/yaw angles can be available to the host processor directly.
With respect to the applications such as 3D gaming, indoor pedestrian dead reckoning, etc., 9-axis or 10-axis sensors are required. In the future, such MEMS sensors and the programmable microcontroller will be combined into a single package. The wireless link and some other sensors may be integrated in the same package too.
Embedded features and sensor integration will determine the future applications of MEMS sensors. Multiple sensors with lower power consumption and a low-cost microcontroller, all in one package, will appear soon.
Solid State Technology, Volume 55, Issue 2, March 2012