MEMS Motion Sensors: The Technology Behind the Technology
MEMS accelerometers and gyroscopes are all the rage in portable design, putting the 'smarts' in smart phones and a new level of fun in gaming consoles. But exactly what are they, how are they made and how do they work?
Despite the fact that MEMS accelerometers have been built into automotive airbags since the mid-90s, few people were aware of their existence until 2006 when the Nintendo Wii game consoles started taking over their living rooms. MEMS motion sensors are now widely used in automotive electronics, medical equipment, hard disk drives, and portable consumer electronics. Today a smart phone can hardly be called ‘smart’ if it doesn’t include a MEMS accelerometer, gyroscope and possibly a compass, too. A small niche product five years ago, MEMS sensors now constitute a multi-billion dollar industry.
So what exactly are MEMS motion sensors and how do they work?
MEMS Motion Sensors
Form follows function, and there are several different types of MEMS motion sensors, each with unique construction and best suited to a particular range of applications.
Single-axis accelerometers (Figure 1) detect a change in velocity in a given direction. They are almost universally used to inflate automotive airbags in the event of crashes. They are also used as vibration sensors to detect bearing wear in machinery, since vibration can be thought of as acceleration and deceleration happening quickly in a periodic manner. Analog Devices, Freescale and Bosch Sensortec all make single-axis MEMS accelerometers that are widely used in these applications.
Figure 1: Analog Devices ADXL150 single-axis accelerometer
Two-axis accelerometers add a second dimension, which can be as simple as detecting tilt by measuring the effect of gravity on the X-Y axis of the accelerometer. Accelerometers come in low-g and high-g sensing ranges, where low-g typically means less than 20x the force of gravity when the measuring body is at rest and high-g can range as high as 100x. Low-g MEMS accelerometers are used in handheld devices; high-g ones find a place in industrial, military and aerospace applications, where the g-forces are in excess of what humans could either generate or withstand.
Three-axis accelerometers can detect motion in three different directions. They widely used in mobile devices to incorporate tap, shake and orientation detection, all of which can result in different actions on the part of a cell phone.
The Freescale MMA7660FC 3-axis accelerometer (Figure 2) targets handsets by incorporating a range of user-programmable interrupts and sample rates in a small footprint (3 x 3 x 0.9 mm) DFN package. The MMA7660FC communicates 6-bit X-, Y- and Z-axis information with the processor over an I2C interface, eliminating the need for an A/D converter. The device draws as little as 47 µA in active mode at one sample per second; 2 µA in standby mode; and 0.4 µA in off mode.
Figure 2: Freescale MMA7660FC 3-axis accelerometer block diagram
Accelerometers measure linear motion, so they're found in applications that measure acceleration, vibration, shock, and tilt. Gyroscopes on the other hand response to rotation, which is a measure of angular motion. They're basically three-axis inertial sensors. Multi-axis MEMS gyroscopes are often embedded along with three-axis accelerometers in inertial measurement units (IMUs).
STMicroelectronics makes the L3G4200D (Figure 3), a three-axis digital gyroscope with only one sensor—a complete inertial measurement device that does away with the need for a separate accelerometer. Gyroscopes have traditionally measured movement around three axes with three sensors—one each for pitch, yaw, and roll; the STMicro device uses a single sensing structure to track all three angular motions. The L3G4200D comes preset with one of three sensitivity levels, which allows the device to trade speed for resolution.
Fabricating MEMS Devices
Figure 3: STMicroelectronics L3G4200D
3-axis digital gyroscope
MEMS sensors are silicon microstructures that are fabricated in much the same way as semiconductors. There are two basic process technologies: bulk micromachining and surface micromachining.
In bulk micromachining single-crystal silicon is etched to form three-dimensional MEMS devices. This is a subtractive process in which the silicon is etched away to form the appropriate microstructure. Bulk micromachining was used to produce the simple motion sensor shown in Figure 1.
Using the same process MEMS pressure sensors are made by etching away certain areas on a silicon wafer to form a diaphragm, which is then bonded to another wafer to form a sealed cavity. Piezoresisters are implanted in the diaphragm; any change in pressure causes the diaphragm to move and the resistance to change, making pressure measurement possible. This is how tire pressure monitoring systems work.
In surface micromachining MEMS sensors are formed on top of a silicon wafer using deposited thin-film materials. Sacrificial layers are deposited and then removed to form the mechanical spaces or gaps between the structural layers. For complex sensor microstructures Analog Devices uses thin polysilicon films; STMicroelectronics and Bosch Sensortech use thick polysilicon films; and Invensense uses single crystal silicon (SCS).
A single-axis accelerometer consists of a movable cantilever beam moving between two capacitive plates, causing a measurable change in capacitance that indicates the amount of movement. Adding an identical perpendicular structure yields a two-axis accelerometer; a third out-of-plane device adds the third axis. In practice most three-axis MEMS accelerometers use bottom plates on a substrate for out-of-plane sensing, posing a trade-off between voltage and sensitivity.
Most surface-machined MEMS sensors use capacitive transduction to convert the mechanical movement of the cantilever arm to an equivalent electrical signal. As the cantilever arm moves between two capacitive plates, the change in gap causes a change in capacitance, which is the electrical equivalent of the input mechanical stimulus. The resulting analog signal is typically digitized and passed to the control circuitry in which the sensor is embedded.
So far we’ve focused on the mechanical side of MEMS devices. But MEMS sensors are usually integrated with electronics to provide a fully functional sensor package. This is a decidedly nontrivial task and an area of ongoing research.
Silicon in Plastic (SIP)
Figure 4: MEMS pin joint
(courtesy Sandia National Laboratories)
First-generation MEMS sensors were silicon-based structures that occasionally included some analog implementation on the same chip. Second-generation MEMS sensors added an A/D converter following a gain stage. Third-generation devices added digital signal and temperature compensation. Current or fourth-generation MEMS sensors further integrate all of that and add on-chip memory.
MEMS devices are typically made from the same single crystal silicon used in semiconductor manufacture. Silicon can be micro-machined to produce very accurate three-dimensional structures on a micrometer scale (Figure 4). Silicon has a yield strength and elasticity equal to that of steel, making some very complex, tiny machines such as micro-motors and micro-pumps possible.
Single crystal silicon, like the single crystal quartz used in watches, also has excellent mechanical resonance properties. This can be both a blessing and a curse. On the plus side a small voltage applied to the cantilever of a MEMS accelerometer will set it vibrating at its resonant frequency. Any motion will cause a change in the rate of vibration, which in turn can be measured simply and accurately with little concern for external electrical disturbances.
On the downside, any shock to the cantilever can set it vibrating at its resonant frequency, rendering measurements inaccurate until this oscillation damps out. In addition a rapidly vibrating cantilever generates parasitic effects in plastic packaging, causing the package to vibrate along with the sensor. This problem long relegated MEMS accelerometers to expensive and bulky metallic/ceramic packages, which had no place in consumer devices. In 2005 Benedetto Vigna’s team at STMicroelectronics managed to solve that problem, producing an inexpensive, plastic-packaged accelerometer that launched the Wii game console in 2006.
Figure 5: Freescale HARMEMS vs. Poly-Si MEMS approach
Freescale takes a different approach to dealing with MEMS sensor vibration problems (Figure 5). In poly-Si MEMS accelerometers a 3 µm cantilever moves between two fixed elements of the same size. Freescale’s high aspect ratio MEMS (HARMEMS) architecture substitutes a 25 µm movable element designed to attenuate the sensor’s resonant frequency and thus any resulting parasitic oscillations. Freescale claims this approach makes for more robust accuracy in automotive safety applications.
System in Package (SIP)
There are two basic approaches to packaging MEMS sensors along with their associated electronics: monolithic integration and two-chip hybrid packaging. Analog Devices uses the monolithic integration approach, though the two-chip hybrid packaging approach is dominant in commercial products.
Figure 6: 3D integrated MEMS platform technologies at KTH-MST
Figure 7: Two-chip hybrid packaging
Monolithic integration involves building the MEMS layer on top of the IC wafer containing the associated electronics. This is a tricky process that is done differently by each manufacturer. The KTH-Royal Institute of Technology (KTH-MST) in Stockholm, Sweden has pioneered research in 3D approaches to monolithic integration (Figure 6). While the advantages of monolithic integration are attractive, achieving it is quite challenging. 3D integration is usually based on wafer bonding. Process stability, yield, and wafer alignment—particularly of inner-chip vias with high aspect ratios—are all critical, particularly with multilayer devices.
Two-chip hybrid packaging is far easier to achieve. Instead of integrating the MEMS layers into or on top of the IC wafer, the sensor die is created separately and then that wafer is either bonded directly on top of the IC wafer or placed next to it on a common substrate (Figure 7). This approach is simple to implement and works well at high speeds, though it’s usually limited to bonding or packaging just two or three wafers together. In the case of MEMS sensors, this isn’t a problem.
There’s an App for That
While MEMS sensors quietly proliferate in medical, industrial and military applications, the real market driver for them is consumer electronics. According to research firm iSuppli, revenue in 2011 for MEMS sensors and actuators used in various consumer and mobile devices will reach $2.07 billion, up 26.2 percent from $1.64 billion in 2010. The five-year market prospects call for growth by a factor of nearly three to $3.71 billion in 2014, up from $1.13 billion in 2009—a solid compound annual growth rate of 23.6 percent during the time period.
While this article focused on MEMS sensors, there’s a whole range of so-called emerging MEMS devices—a broad array that includes such items as radio frequency (RF) MEMS switches and varactors, timing devices, and autofocus and zoom actuators that are rapidly coming online. According to iSuppli, by 2014 emerging MEMS devices will account for 39 percent of overall consumer MEMS revenue, compared to just 10 percent in 2010.
MEMS is one hot technology. Stay tuned for more exciting developments.