MEMS INERTIAL SENSoRS have been around for more than 25 years, from the first prototypes developed in universities to initial product offerings from the likes of Analog Devices, Bosch, Motorola (now Freescale) and STMicroelectronics. Driven by the Defense Advanced Research Projects Agency, with the promise of ubiquitous inertial navigation, MEMS sensors saw their first commercial success as collision detectors in vehicle airbag systems.
Fast-forward 20 years. Airbags that started on high-end cars are now available on every vehicle. The use of sensors in cars has expanded into all manner of vehicle dynamics. Improvements in the cost, size and power of current sensors, along with expansion of the vendor base, have allowed MEMS sensors to spill over into consumer electronics such as game controllers and smartphones.
MEMS inertial sensor employing variable capacitor transduction.
Source: HP Labs
Their performance, however, remains largely unchanged. The systems are a bit more digital now, but factors such as noise floor and stability, which are key to navigation, have not seen the same improvements as cost, size and power usage.
It is time to examine why MEMS inertial sensors have failed to live up to their initial promise, and to propose a different approach that could jump-start those advancements. MEMS fabrication technologies, such as high-aspect-ratio etching, wafer bonding and packaging, have all seen dramatic improvements from the first university prototypes. A new, nano-based approach that could dramatically move the needle toward realizing a MEMS-based inertial measurement unit suitable for consumer electronics.
Hewlett-Packard's 25-plus years of nanofabrication experience was not in sensors, but instead focused on creating printheads with ever-increasing nozzle density, shrinking droplet size and improved power efficiency. Using that knowledge, HP took out a clean sheet of paper and designed a MEMS-based storage device. In the process, we discovered a technology platform with all the attributes needed to create a new generation of MEMS inertial sensors.
The storage device sought to miniaturize a CD-RW-like system onto a chip. The rotating disk and laser were replaced by an XY positioner, dubbed the micromover, and an array of field-emitting electron guns. Both devices use phase-change media to store data bits.
The program developed several unique MEMS technologies, but the micromover, created with high-aspect ratio etching and wafer bonding, was the clean sheet needed to transform MEMS inertial sensors. By taking the precision XY actuator and running it backward, the CD-RW-like system-on-chip became a high-performance sensor.
It's no accident that the micromover made a great sensor. Good storage devices and high-performance sensors share the same required attributes:
• Thermal stability. The most critical parameter of any storage device or inertial sensor is its ability to handle changes in temperature without a loss of performance. Thermal stability affects the ability to quickly find data bits or integration errors from drift in the acceleration output.
To ensure thermal stability, designers can use an all-single-crystal silicon construction with a minimum of metal and dielectric layers to minimize stress and ensure the whole device changes cohesively with temperature fluctuations.
• Large proof mass. For storage devices, capacity is all about area for storage media. To increase capacity, you build a larger micromover. In the sensor, increased mass is the key to lower thermomechanical noise.
Nano is not always your friend in the business of MEMS inertial sensors. If you are working with “the world's smallest inertial sensor,” you will also have to factor in compromises to address issues like thermomechanical noise, which will limit the sensor resolution. But it is possible to achieve a small-form-factor/high-performance balance.
• Off-axis isolation. For the storage device, it is critical that the gap between the media and the electron guns be controlled precisely. The media must be kept in the focal plane of the electron beam to allow reading and writing of the data bits.
High-aspect-ratio through-wafer etching is used to create a flexural suspension with out-of-plane stiffness that is high enough to control the gap. This translates into off-axis isolation for the inertial sensors, as well as increased mass from the thicker proof mass and flexures.
• Dynamic range. The farther the micromover can travel, the higher the data volume accessible by each read-write electron gun. Increasing the travel range reduces the number of guns necessary and, as a result, reduces system complexity and cost as the number of parallel channels decreases.
To increase the travel range, we developed an electrostatic surface electrode arrangement that was a revolutionary step from the interdigitated finger electrodes previously used in MEMS actuators. The approach enables precise gap control with wafer bonding and dimension control with lithography, letting engineers achieve a new range of force density for an actuator or capacitance change per acceleration in sensor mode.
The biggest design advantage, however, is that the surface electrodes are not travel-limited, as they are in the finger-based arrangement. Thus, the new approach provides a larger dynamic range in addition to high-resolution sensing. For example, the micro-g accelerometer is capable of a measurement range exceeding 10 g.
Increased proof mass reduces thermal vibration noise for more-precise acceleration measurements.
Source: HP Labs
Testing has shown that using the new technology platform can result in orders-of-magnitude improvements in noise floor, dynamic range and stability, while maintaining the traditional MEMS advantages of cost, size and power. The increased flexibility of the system lets engineers build accelerometers and gyroscopes that can measure all six axes of motion necessary for navigation–XYZ, roll, pitch and yaw–in a single planar chip. Single-chip integration ensures precise alignment of axes, reducing packaging costs over other available technologies.
With the accelerometer technology in place, designers have an ultra- sensitive, powerful sensor that achieves the size and performance levels required for several crucial application areas in which factors such as lower noise floor and large dynamic range are the enabling features.
For instance, design engineers working with structural health monitoring systems require a sensing system that can deliver real-time measurements of the motion and mode shape of a bridge. A large bridge, such as the Golden Gate, could use hundreds or even thousands of sensor nodes equipped with accelerometers remotely monitoring vibrations to track the aging of the structure and signal an alert in the event of a variation from the normal data output. The sensor nodes need to be small, rugged, low cost and low power to minimize maintenance of the system once installed on the structure.
Current MEMS inertial sensors have been used in several bridge monitoring studies. While the devices meet the size and power requirements for the applications, we believe that providing higher sensitivity via the sensor will enable more accurate imaging of structural motion. Using sensors with lower noise (for higher sensitivity) and improved stability (for higher resolution of the low-frequency vibrations typical of large structures), a sensor node can be optimized for monitoring structural soundness.
This type of high-performance sensor networking system could also be used in geophysical applications to monitor motion during a seismic event such as an earthquake. The information captured in real-time during the quake could be used to quickly determine structural safety following the event.
The continuing evolution of MEMS technology offers significant impacts on how we obtain information from and interact with our environments. Nanosensing technology is at the point of enabling the performance, size and cost points required for use in the widely deployed networking sensor systems that will drive these types of critical systems.
Peter G. Hartwell is a senior researcher for information and quantum systems at HP Laboratories, where he leads the MEMS team. He holds a bachelor's degree in materials science and engineering from the University of Michigan, and a Ph.D. in electrical engineering from Cornell U.