In a conference room at Analog Devices (ADI), Howard Wisniowski holds a demo board a little bigger than a commemorative stamp about a meter above the table top. An ADI motion sensor and associated circuitry are on the board. Wisniowski drops the board into his other hand.
As soon as the board starts to fall free, the motion sensor detects a change in acceleration. Before the board reaches Wisniowski’s lower hand, an LED flashes red and a tiny transducer on the board squeaks out an SOS in Morse Code. The speed of that response, reoriented to the horizontal plane, is just what you would want in a notebook disk-drive head assembly or an automotive airbag sensor.
Those are just two of the new applications areas opened up by the economies of scale, high sensitivities, and small footprints made possible by microelectromechanical-systems (MEMS) manufacturing technologies adapted to CMOS process flows.
THE INSIDE STORY
The sophistication in MEMS accelerometers lies partly in the electronics and partly in the geometry of the mechanical configuration. Accelerometers can be fabricated and packaged to measure acceleration in a single plane or in two or three orthogonal planes. Conceptually, the acceleration-sensing portion generally comprises a “proof” mass at one end of a cantilever beam.
The deflection of the system of multiple proof masses and beams under acceleration is often measured by sensing a change in capacitance between a set of fixed beams and a set of deflecting beams—somewhat analogous to a macro-scale variable capacitor. Since many capacitive transducers have a nonlinear capacitance versus displacement characteristic, the electronics in the sensor are called upon to convert the signal to a linear output. Alternative sensing elements may be piezoelectric, rather than capacitive.
Important accelerometer datasheet characteristics include bandwidth and resonant frequency, noise floor, cross-axis sensitivity, drift, linearity, dynamic range, shock survivability, and power consumption. Generally, resonant frequency is several times higher than the upper bandwidth limit. Bandwidth and sensitivity tend to be inversely related.
In addition to the usual noise sources in electronic devices, MEMS sensors are so small that Johnson noise, caused by Brownian effects on the proof mass, can have a significant effect. A nice derivation of Brownian effects can be found in “Sensors— An Overview of MEMS Inertial Sensing Technology” at www.sensorsmag.com/sensors/content/printContentPopup.jsp?id=334974.
So far, we’ve been considering linear accelerometers, which have a huge market in transportation applications, particularly airbag-related deceleration sensing. But a large market also exists for MEMS angular accelerometers in disk drives, where they compensate for angular shock and vibration. Unlike their linear cousins, these devices locate their centers of gravity at the centroid of the support springs, making them sensitive to angular acceleration.
Acceleration, vibration, shock, and tilt relate to linear rate motion. Rotation is a measure of angular rate motion. This mode differs from the others because rotation may occur without a change in acceleration. To understand how that works, picture a three-axis inertial sensor.
Say that the sensor’s X and Y axes are parallel to the Earth’s surface. The Z axis is pointing toward the center of the Earth. In this position, the Z axis measures 1 g. The X and Y axes register 0 g. Now rotate the sensor to move only about the Z axis. The X and Y planes simply rotate, continuing to measure 0 g while the Z axis still measures 1 g.
That’s why MEMS gyroscopes are used to sense this rotational motion. Because certain end products must measure rotation in addition to other forms of motion, gyroscopes may be integrated in an inertial measurement unit (IMU) that embeds a multi-axis gyroscope and multi-axis accelerometer.
For a good heuristic video on the complementarities of accelerometers and gyros, check out www.invensense.com/support/videolibrary.html on InvenSense’s Web site. (The InvenSense site also has some excellent white papers on topics like image stabilization and MEMS gyroscopes.)
Accelerometers are all about displacement and vibration in a plane. With MEMS gyros, it’s about displacement caused by Coriolis force. While it may have nothing to do with water going down the bathtub drain, the Coriolis force does work on smaller scales than hurricanes, and it’s demonstrated on a more moderate, though still far from microscopic, scale in those Foucault pendula that every science museum seems to have.
Assuming that all you remember about Foucault’s pendulum is that it has something to do with the rotation of the earth, here’s the short explanation. Suppose you have a vibrating mass particle (the pendulum ball) moving at resonance with velocity v0 cos(Ot) that’s fixed to a body (planet Earth) rotating at rate Oin. The Coriolis effect induces a time-varying acceleration at the same frequency as the driving acceleration, but at right angles to the velocity vector of the mass particle. That is, the Coriolis acceleration is a cross product: a(t) = \\[ 2Oin × v0\\] cos(Ot).
Now, mentally change the huge Foucault pendulum to a vibrating tuning fork and you have a similar effect (Fig. 1). The tuning fork’s normal vibration mode is in one plane and the Coriolisinduced displacements are in another plane that’s orthogonal. Shrink that to MEMS size, drive the tuning forks with an external signal, use separate tuning forks for three axes, and you have the basic concept of a three-axis MEMS accelerometer.
Continue to page 2
The “basic concept,” of course, ignores the challenges of actually manufacturing the device, cross-coupling vibrations from one axis to the other, calibration, thermal issues, and so forth. And there’s no need to make the vibrating element look like a piano tuner’s tool. Imagine what you could do with a circular structure that vibrates like the mouth of a bell or a wine glass (Fig. 2). There are lots of patents on MEMS structures and many clever ways of adapting semiconductor process flows to manufacturing these devices.
BREADTH OF APPLICATIONS
People tend to think of accelerometers in terms of automobile airbag deployment. But really, since movement and position changes are accompanied by acceleration, it’s common to use MEMS accelerometers to detect events that are less dramatic than cars running into each other.
Whenever a device is picked up and put down, the accompanying change of acceleration can be detected and used to generate an interrupt that powers certain device functions on and off, keeping some active while putting others into power-saving sleep states. Think of a handheld device that turns off its backlight until it’s picked up. (Of course, the movement sensor had better consume a lot less power than the backlight!)
More dramatically, a year or so ago, I wrote about portable radios for first responders that signal automatically when the person carrying them stops moving for a certain length of time— for instance, a lone firefighter disabled in a burning building (see “P25 Handhelds Incorporate High-Velocity Human Factors Design”). Or on a battlefield, you wouldn’t want the enemy picking up a radio from a dead or wounded soldier and using it to obtain tactical intelligence, so the radio can be programmed to require re-authentication before permitting user access.
Wisniowski described new defibrillators for use in public places. These devices are designed to help relatively inexperienced rescuers deliver CPR when electric heart stimulation fails. When that happens, Wisniowski said, “A less experienced rescuer might not compress the patient’s chest enough for effective CPR. Accelerometers embedded in the AED’s chest pads can be used to give the rescuer feedback on the proper amount of compression by measuring the distance the pad is moved.”
When Electronic Design publishes a story about energy harvesting, the most common application is vibration monitoring to assess the condition of mechanical systems like industrial pump motors, railcar wheel bearings, and highway bridges. Generally, one reads about the means of using the energy of the vibrations being monitored to power the microcontroller and the mesh-radio node, but I rarely consider the source of the raw data.
Yet that’s a key part of the system. Very small MEMS accelerometers with very wide bandwidth are required to capture an accurate enough profile of the normal vibration baseline and the aberrations. It would ultimately provide enough diagnostic information to allow intelligent analysis of potential time to failure.
So far, we’ve considered displacement and vibration. Shock impulse events are another source of accelerations. Probably the widest use of such sensing is in notebook disk drives. Interestingly, with disk drives, it isn’t the shock itself we want to detect. At that point, it would be too late to do something about it.
As Wisniowski’s “SOS” demonstration suggests, even before the shock itself, it’s possible to detect the changes in g-forces that indicate the notebook has been dropped, which are precursors to damaging shock associated with hitting the floor. During the milliseconds that elapse between the onset of zero-g conditions and the notebook’s impact, the system can order the disk-drive head to be parked.
The Apple iPhone and Nintendo Wii have accustomed us to the use of accelerometers and gyros for gesture recognition—taps, double-taps, and shakes that activate features and adjust operating modes. In addition to adding coolness to the product, providing gesture input has other benefits, Wisniowski observed.
Button-free designs have other advantages in lower system cost and higher ruggedness in products such as underwater cameras. Tap interfaces also are appropriate in wearable and implantable medical devices.
The Wii game control has also introduced a wide audience to tilt sensing. Beyond gaming, tilt sensing offers interesting potential in industrial applications. In these cases, operating a device one-handedly can leave the other hand free for hanging on to a vehicle transiting uneven ground or for control of a bucket or platform. Here, you would use a three-axis accelerometer to detect slow changes in inclination in the presence of gravity, interpreting that as a twist or tilt in one direction or another.
Looking at more prosaic applications than Wii-like control of industrial equipment, lots of jobs would involve tilt-sensing capability. Examples include adjusting industrial weigh scales and pressure for proper orientation.
At the other end of the complexity spectrum, the latest IMUs combine a multi-axis accelerometer, a multi-axis gyroscope, and a multiaxis magnetometer. ADI’s six-degrees-of-freedom IMU provides fine resolution in medical imaging and surgical instrumentation.
In mid-2007, ADI broke new ground with the ADIS16355 IMU. It combines three axes of angular rate sensing and three axes of acceleration sensing, bringing 50 times greater accuracy than other off-the-shelf inertial sensors. It also came pre-calibrated, meaning that data out is accurate regardless of operating temperature. The product designer needn’t incorporate a table of correction values in system code.
At introduction, in 1000-unit lots, the full-range temperaturecalibrated version cost $359, and the room-temperature calibrated version cost $275. The device comes in a cube measuring just shy of an inch per side, with a little extra space required for mounting feet and a flex circuit with a connector on the end.
Continue to page 3
Obvious target applications are vehicle-mounted cameras and antennas, commercial aircraft guidance units, robotics, and prosthetics. Another important area is inertial backup when GPS signals are lost. That’s significant not only in aviation, but also in commercial-vehicle fleet operations and automated harvesting equipment on mega-farms.
Specs were impressive, with 14-bit precision. Output and control is via a simple serial peripheral interface (SPI) port. Each gyro has a ±300°/s dynamic measurement range, and each accelerometer has a ±10-g measurement range. And although their maximum dynamic range is ±300°/s, the IMUs provide ±75°/s and ±150°/s ranges as well.
Each sensor’s signal-conditioning circuit has an analog bandwidth of approximately 350 Hz. The IMU provides a Bartlett Window finite impulse response (FIR) filter with programmable step sizes for additional noise reduction on all of the output data registers.
In addition to the calibrated motion measurements, the IMU measures power supply and temperature, as well as provides an auxiliary 12-bit analog-to-digital converter (ADC) channel. This output data updates internally, regardless of user read rates. Output data can be either 12 bits or 14 bits long.
An auxiliary 12-bit successive-approximation ADC makes it possible to digitize other system-level analog signals. Furthermore, an auxiliary 0- to 2.5-V output digital-to-analog converter (DAC) provides a 12-bit level-adjustment function.
About six months before the 16355 hit the market, ADI had introduced the ADIS16209 dual-axis MEMS inclinometer and accelerometer for industrial applications (see “Tiny Dual-Axis MEMS Inclinometer Simplifies Industrial Measurements”).
Late in 2008, we saw the four-degree-of-freedom ADIS16300 and six-degree-of-freedom ADIS16405 IMUs with 14-bit gyroscope featuring digital range scaling; ±75°/s, ±150°/s, and ±300°/s settings; a tri-axis, 14-bit, ±5-g digital accelerometer; and 180-ms response time (Fig. 3). In addition, they provided factory-calibrated sensitivity, bias and axial alignment, digitally controlled bias calibration, and a digitally controlled sampling rate up to 819.2 samples/s. (An external clock allows sampling up to 1200 samples/s.)
Also released in that timeframe was the ADIS16209 dual-mode inclinometer. It delivers either dual-axis horizontal operation of ±90° or single-axis vertical operation of ±180°. It operates from a 3.3-V power supply and communicates via an SPI bus.
When it was announced, ADI said its dimensions of 9.2 by 9.2 by 3.9 mm made it 100 times smaller than other available products and that it cost one-tenth the cost of functionally equivalent competitive units. As noted earlier, multiple applications are medical. Ultrasound, mammography, and X-ray equipment all need precision and accuracy in scanner alignment. The devices are being used in hip and knee surgical procedures as well.
Those kinds of medical applications are also one target of the six-degree-of-freedom 16405 IMU, with its tri-axis magnetometer sensor for heading sensing. ADI says that it, too, costs up to 10 times less than competitive products.
Last March, ADI announced the ADXL346 digital three-axis iMEMS smart motion sensor as part of its family of small, powersipping smart motion sensors for portable devices. It operates at primary supply voltages down to 1.8 V and comes in a 3- by 3- by 0.95-mm package. Furthermore, it measures both dynamic acceleration (resulting from motion or shock) and static acceleration (such as gravity, which allows it to be used as a tilt sensor).
To save power, it buffers up to 32 sample sets of X-, Y-, and Z-axis data in a first-in first-out (FIFO) arrangement, enabling the host processor and other power-hungry peripherals to go into a sleep mode until needed. Bandwidth is selectable from 0.1 to 1600 Hz, allowing tradeoffs between responsiveness and battery life. Power consumption ranges from less than 150 µA at 1600-Hz bandwidth down to 25 µA under 10 Hz.
The ADXL346 measures dynamic acceleration with ±2/4/8/16-g user-selectable ranges and includes built-in orientation sensing via simple register reads. Special sensing functions with userprogrammable thresholds include inactivity, tap/double-tap, and free-fall sensing. Pricing for the ADXL346 is $3.04 per unit in 1000-unit quantities.
Earlier in the year, ADI unveiled the ADXL345 three-axis accelerometer, claiming an 80% power savings compared to competing three-axis sensors. Power-saving design elements include the low single-cell operating voltage and FIFO arrangement like the one described before, which offloads the task of responding to a change in movement or acceleration from the host processor. Also, the output data range scales from 0.1 Hz to 3.2 kHz, allowing portable-system designers to precisely allocate power for specific system functions. Pricing is $3.04 in 1000-unit quantities.
ADI, of course, isn’t the only manufacturer of MEMS motion sensors. Freescale Semiconductor’s MMA745xL three-axis digital- output accelerometers for mobile devices support tilt scrolling in all directions, gaming control, gesture recognition, and tap to mute. They also support theft protection, freefall detection, and GPS backup applications.
STMicroelectronics’ LIS302DLH 16-bit three-axis accelerometer suits motion sensing, orientation awareness, freefall detection, and vibration monitoring (Fig. 4). At 0.75 mm tall, it is the market’s thinnest device, according to the company. (Otherwise, it shares the 3- by 5-mm footprint of other devices in ST’s Piccolo MEMS family.) It outputs acceleration data up to ±8 g via a serial peripheral interface (SPI) bus. Currently sampling, production pricing is $1.35 for orders over 10,000 pieces.
Continue to page 4
To complement its acceleration sensors in applications ranging from gaming and remote-pointing to car navigation and compensation for camera shaking, STMicroelectronics also recently introduced a family of 15 single-axis and multi-axis MEMS gyroscopes. The family, which comprises a wide 30- to 6000-dps (degrees per second) full-scale range, includes single-axis (yaw) and two-axis (pitch-and-roll, pitch-and-yaw) devices.
Either configuration can provide two separate outputs for each axis at the same time—an unamplified output value for the general detection of angular motion and a four-fold amplification for high-resolution measurements. High-volume unit pricing is $2.50.
For applications such as disk-drive protection in handhelds, STMicroelectronics has announced a three-axis accelerometer with absolute analog output that operates at low voltages—2.16 to 3.6 V. The LIS352AX is insensitive to battery power-supply voltage variations and demonstrates high stability over a wide temperature range for both zero-g offset and sensitivity. Builtin self-test makes it possible to verify sensor functioning after board assembly. Volume pricing is $1.30 each.
In March, Cornell University spinoff Kionix introduced the KXTF9 tri-axis accelerometer with a new interface it calls “Directional Tap/Double-Tap.” It creates up to 12, unique, tap-enabled commands for end-use developer-specified functions. Directional Tap/Double-Tap detects quick, light taps or double taps on any of the six faces of an object.
According to CEO Greg Galvin, “A single tap to the face of a cell phone could send the call to voicemail or silence the ringing. A tap to the left could enable the navigation functionality. A doubletap on the bottom could provide a transition to Internet access.” Other features include a user-programmable output data rate (ODR), selectable 8-bit or 12-bit resolution, user-selectable 2-, 4-, and 8-g g-ranges, and a digital high-pass filter with a userselectable cutoff frequency. The device operates at supply voltages from 1.8 to 3.6 V dc.