The silicon micromachined inertial sensor that deploys your automotive airbags can't simply be dropped into your laptop for free-fall detection without some do-differentlys. The differences go well beyond the application spaces of medium/high-g and low-g that automotive sockets sport. In fact, the automotive and consumer markets present conflicting fundamental demands. Building a bridge between them means giving reconsideration to design, test, space, quality, time-to-market, front- and back-end assembly—and price.



Some Overlap, More Differences

Automotive and consumer inertial-sensing application domains overlap in the realm of single-digit g's. Automotive accelerometer sockets address safety requirements and range from ~1 g (e.g., tilt, dynamic suspension) to 250 g and above (impact or collision detection).

Consumer applications are designed for more benign acceleration environments; their value lies in product protection and feature enhancement.

As you can guess, the consumer OEM sector is more diversified and fragmented than the automotive arena. More and more consumer products, such as cell phones (in particular those supporting games and including digital video capability), portable hard-disk-drive devices, video games, car navigation systems, and digital cameras have begun adopting inertial sensors to monitor tilt, flag motion events, and detect shock, vibration, and position. GPS-based products, pedometers, and a host of devices that could benefit from anti-theft features can all take advantage of the capabilities provided by inertial sensors.

Moreover, the automotive and consumer markets differ in how they rank the relative importance of performance, reliability, and cost for low-g applications. Safety-critical automotive apps understandably demand the highest in quality and reliability; the most important consumer market constraints are typically cost, time-to-market, and low power consumption. Now let's look at some of the finer points of this distinction.

Price and Time-to-Market Call the Shots

In the consumer market, low price is usually second only to supplier selection—sometimes to the extent that features are slashed until the price can meet the customer's target cost. The conventional methods that semiconductor vendors rely on to reduce cost—smaller die size, shorter test time, improved yields, low-cost packaging, and high-volume production—are still used. But now heightened emphasis must be given to shrinking the time and costs of development, as well as rethinking design engineering and all facets associated with touch labor. Fortunately, it's possible to repurpose technologies developed for automotive products and still achieve shorter design cycles, high quality, and low cost.

The special case of inertial sensors with a two-chip (transducer and signal conditioning) approach is a good example. IC technology developed for products such as analog devices or microcontrollers can be reused for signal conditioning. A two-chip technique is also conducive to shorter design cycle times since transducer and IC design can be done in parallel, resulting in a shorter effective wafer process time.

Time is Always Money

Development, test, and burn-in during production are the three biggest cost contributors in inertial sensor qualification. In accordance with accepted practice, burn-in is removed from the consumer production flow. As for test, cost is further reduced by a judicious choice of parameters at which operational corners are tested. This is usually based on a vendor-customer agreed-upon list of critical parameters. The remaining electrical parameters are guaranteed by design, an approach justified by using enhanced characterizations with previously established very high parametric Cpk (process capability) values over extended qualification regimens. The high quality and reliability pedigree of automotive readily comes across.

Because consumer parts are usually specified with wider operational windows as compared to their automotive "cousins," the cost of this approach can be zero. In other cases, the cost is minimal when additional test intervals are folded into pre-existing characterization and qualification activities. This dramatically compresses new product introduction cycle time, as consumer product buyers usually accept the pre-existing data from the cousin parts or part components, while they look to their own module-level qualification and the qualification of record. This joint working relationship is mutually beneficial.

Time-to-market can be shortened and power consumption can be reduced also by tailoring or reusing automotive-based platform elements, building blocks, and packaging. The performance demands for consumer products are less stringent than those for automotive applications. Circuits can therefore be simpler, which helps reduce the die size and probe and final test time, and improve yield as well. The resulting consumer package is not only low-cost (e.g., plastic), but also smaller and thinner for portable products.

Motion Recognition and New User Interface

Accelerometers now enable smarter user interfaces and controls, such as by detecting a combination of shock, position, tilt, and motion (Figure 1). For example, menu selection can be initiated by an accelerometer that senses small shocks from finger tapping. Menu scrolling can be initiated from a tilt measurement where inclining the device forward, backward, or side to side moves scroll bars up or down, left or right. In the past year, consumer electronic devices have incorporated these basic motion detections to initiate functions usually accessed through pressing multiple buttons.

Figure 1. Accelerometer output during the motion of picking up a phone and bringing it to your ear
Figure 1. Accelerometer output during the motion of picking up a phone and bringing it to your ear

A motion recognition system can be implemented with varying levels of complexity. The simplest implementation would include a 1-, 2-, or 3-axis accelerometer and a low-cost microcontroller. The microcontroller waits for an acceleration value or a sequence of accelerations that matches a sequence or logic pattern stored in memory. When a preprogrammed motion or acceleration limit is detected, the microcontroller executes the respective command (logic output with an I/O pin) or control sequence (signal with a SPI or PWM output). The processor could be an 8-bit microcontroller such as the MC68CH908QG8 for simple motion detection algorithms (Figure 2).

Figure 2. A basic schematic showing how to interface the accelerometer to an 8-bit MCU
Figure 2. A basic schematic showing how to interface the accelerometer to an 8-bit MCU

As more complex motions are programmed for recognition, the algorithms become more sophisticated and require more ROM and greater processing capability.

Motion Detection

For motion detection, the accelerometer must accommodate many ranges of hand movements (Figure 3 ). Small hand motions, such as those controlling a computer mouse or scrolling a menu, would require 1–2 g's. Larger gestures, such as when a person is moving his or her hands while talking, need 3 or more g's, and any quick change in acceleration such as fist-shaking calls for 4 g's or more. Freescale's low-g accelerometers with a g-select option address this requirement wherein two logic inputs to the accelerometer can be driven by a microcontroller to change the sensitivity and g-range dynamically. For example, the MMA7260Q provides the option to select 1.5, 2, 4, and 6 g levels of acceleration.

Figure 3. G-select enables multiple applications with one accelerometer
Figure 3. G-select enables multiple applications with one accelerometer

Out-of-Balance Detection

The general philosophy concerning home appliances is simple: the less human intervention during operation, the better. Accelerometer-based systems can incorporate a closed loop that determines the machine's status and any corrective measures required. In washing machines, for example, excessive vibration or an out-of-balance condition can occur during the spin cycle. An accelerometer can detect this oscillatory motion by calculating the rotations per minute and the diameter of the rotation. When those values reach a set threshold, the spin cycle would be automatically stopped for the tub to be shaken and clothes redistributed before the machine begins to walk around the room. A more complex solution could determine the current out-of-balance motion and spin the tub at the optimal speed, which is just slower than the point where audible knocking would occur.

Accelerometers designed for appliance applications have fewer size constraints than those intended for laptops, cell phones, and other small electronic devices. Power consumption is less of a concern as well because appliances are not battery operated. And, of course, the accelerometer's current draw (in the milliamp range) is insignificant compared to that commanded by the appliance itself. High reliability, media protection, and cost are, however, of great importance, and even more so, sensitivity in the 6–10 g range for the out-of-balance application.

Freefall Detection

Given the number of handheld devices incorporating hard disk drives now on the market, it is not surprising that there has been a soaring demand for freefall detection capabilities. An accelerometer can sense three different types of freefall, or free fall signatures — linear, rotational, and projectile. A linear fall is defined as a linear translation of an object falling from any orientation, where the orientation does not change during the translation. A rotational fall is a linear translation of an object falling from any orientation, where the orientation changes by the object's rotating on its axis. A projectile fall is a planar translation with two dimensions, vertical and horizontal, where the object is essentially thrown in the horizontal direction while falling in the vertical direction. In any or all of these events, a signal can be sent to the hard drive to park the disk and protect the stored data.

The minimum requirements for a freefall detection system are a 1.5 g accelerometer and an 8-bit microcontroller with 8 KB of Flash memory. This design ensures that all three freefall signatures will be quickly and reliably detected (Figure 4) because there are no other processes running that could take precedence over the freefall detection algorithm. Most end products, however, need a more complex processor such as an MCU for PDAs or a DSP for MP3 players. In these products, the freefall algorithms can be easily added as an interrupt function to the main processor.

Figure 4. The output of a 3-axis accelerometer; during linear freefall, all axes converge to zero g
Figure 4. The output of a 3-axis accelerometer; during linear freefall, all axes converge to zero g

The accelerometers designed for this purpose feature reduced power requirements that can be satisfied with batteries. In addition, the g-select feature permits multiple applications to run simultaneously. One accelerometer can therefore toggle between freefall and a new user interface application. This serves to reduce cost concerns by adding value beyond the accelerometer's protective function.

Virgil P. LaBuda, MS, and Michelle A. Kelsey, BSCS, can be reached at Sensor Products Div., Freescale Semiconductor Inc., Tempe, AZ; 480-413-8869, [email protected], or 480-413-3533, [email protected], www.freescale.com.