MEMS sensors have been around for a long time, but requests from the market for new applications are driving upgrades in the technology. Because of their small size, their accuracy, and reliability, MEMS sensors are a good fit for wearable devices.

Barometric pressure sensors, for example, are ideal for embedding in watches, fitness bands, earphones, or smartphones to support fitness parameters that can sense whether you are walking on a flat area, up an incline, or on stairs. They add a third (Z) dimension to standard two-dimensional (X, Y) navigation devices. In fact, some are sensitive and accurate enough so an emergency call can signal to a rescuer what floor you are on in a skyscraper. Bosch Sensortec has now introduced the BMP 384 robust barometric pressure sensor into the wearables market. It has been ruggedized with a gel-filled package, which is an innovation for a barometric pressure sensor. This makes it resistant to water and various harsh environments. In addition, because of its small size — 2.0 × 2.0 × 1.0 mm — it is easy to integrate into smartphones, wearables, and hearables.

Bosch has also introduced the BHI260AP self-learning AI sensor, which can be taught to track individualized workout routines that include a variety of different movements. And since it has a water-resistant case, it can also track laps, strokes, and rest times for a swim workout.

Inside a MEMS Barometric Pressure Sensor

Figure 2. The water-resistant BMP 384 barometric pressure sensor can be used to track swimming. (Image courtesy of Bosch Sensortec)

The mechanism of the Bosch BMP series of pressure sensors, is a 10-to-20-micron thick membrane in a silicon die. There are four piezoresistive elements in the membrane so that when it bends in response to pressure, the change of resistance of the piezoresistors varies the output of a Wheatstone bridge, thereby producing the pressure signal.

This is a well-proven technology, which has been used in the automotive market for more than 20 years and has been shown to be quite stable even under rugged conditions. It is now entering the consumer electronics industry with a successful track record behind it. Our latest entry to this market is the model BMP384, which is not only robust, but is media-resistant due to its gel-filled package

Accuracy

Relative accuracy is a function of the slope of the output curve; it is the accuracy of changes in altitude, rather than an absolute reading. For the high-performance Bosch BMP 390, it is +/- 0.03 hPa., which is equivalent to +/- 25 cm.

Absolute accuracy is a measure of the maximum error in the readout of the exact pressure. For the high-performance BMP390, it is +/- 0.50 hPa. Other specifications relevant to absolute accuracy include RMS noise: +/- 0.02 hPa; temperature coefficient offset: (25 °C – 40°C, at 900Pa); 12-month stability: +/- 0.16 hPa; and solder drift: <+/- 0.8 hPa.

(Solder drift, although not often mentioned, can be a significant factor. When soldering a device to a printed circuit board, you create a temperature gradient, which can cause a mechanical stress, including bending. This kind of stress on the MEMS sensor can cause changes in the electrical output.)

Navigation

MEMS barometric pressure sensors can be optimized for different applications. For example, the BMP390, designed for accurate elevation measurement, has good enough resolution to measure height changes of less than 10 cm. With a case size of 2 mm × 2 mm × 0.75 mm, it can be easily integrated into a smartphone or watch. Bosch, in conjunction with NextNav LLC (Sunnyvale, CA), has developed an indoor navigation system that uses the BMP390 to provide the indoor z-axis location component for three-dimensional location and positioning, good enough to be used for enhanced emergency calls (E911).

For an emergency call, absolute accuracy is critical. Let’s say you live in an apartment on the 14th floor, for first responders to find you quickly, the z axis information, as well as the x and y, all have to be precise and accurate.

To achieve the most accurate absolute readings, you would have to start with a calibration that takes into account the altitude of your location, for example, whether you are at the seashore or on a mountain top. Then, there are several additional factors that can affect the pressure sensor’s absolute accuracy. For example, the correlation between pressure and altitude varies as the outside barometric pressure changes. This is not generally significant for rapid changes in altitude, but for example, if your phone with its imbedded sensor is in one particular location for 10 hours, the environmental pressure will likely change over that period. So that means it is necessary to bring in an external corrective signal. That can be achieved with sensor fusion, combining information from several different types of sensors, such as inertial and magnetic.

Figure 3. Learning new exercises with the self-learning AI sensor BHI260AP. (Image courtesy of Bosch Sensortec)

Fitness — Pressure Sensing

For tracking workout routines, relative accuracy is fine — you’re only interested in the change in altitude, not the absolute value. Environmental factors, however, are significant for fitness trackers, especially for wearables, such as smart watches, especially if you want to track swimming.

The gel filled BMP384 can be used for fitness tracking, especially when it comes to changes in altitude or fitness in harsh environments like water.

In order for the sensor to operate in harsh environments, the sensing mechanism as well as the embedded ASIC has to be isolated. In the BMP 384 this is done by applying a layer of gel between the diaphragm and the case. It requires special know-how to make and apply a gel with the right mechanical and thermal properties. The gel has to transmit pressure to the membrane, while not blocking it from bending. It also has to conduct heat well enough so that the sensor works over a wide temperature range and finally, it must not stiffen over the working lifetime of the sensor.

Fitness — Artificial Intelligence

The BHI260AP MEMS sensor, which was introduced by Bosch at CES 2021 is designed to be integrated into wrist wearables, such as smart watches and fitness bands; or hearables. It not only contains a six-degree-of-freedom IMU with a 16-bit 3-axis accelerometer and a 16-bit 3-axis gyroscope, but also a 32-bit customer programmable microcontroller. This combination of hardware plus included self-learning AI software supports a very sophisticated fitness device.

Swimming.Bosch Sensortec has developed swimming tracker software for the BHI260AP. Using real-time sensor data from the IMU and the floating-point microcontroller, it can provide both raw sensor data as well as run AI functions that generate relevant results for direct use by an application processor. The built-in motion sensor determines when the user has started swimming, without requiring any action from the swimmer. It then classifies the stroke type from four possible categories — backstroke, freestyle, butterfly, and breaststroke — and records the number of strokes, laps, and any breaks between the laps.

Figure 4. The new self-learning AI sensor is able to learn, personalize, auto-track, and enhance workouts. (Image courtesy of Bosch Sensortec)

Home workout. This “smart” sensor can be trained to be a personal assistant for your home workout — It can recognize and track the details of your personal exercise routine. Although it comes with 15 pre-programmed fitness activities, additional activities can be uploaded.

The sensor is smart enough to recognize new exercises and can adapt to match your own specific workout. By learning from your behavior, it can recognize hundreds of different movements and patterns, not just those that the device manufacturer has pre-programmed.

For each exercise, you can get detailed and instantaneous feedback, for example: the type of activity, the time required, and the number of sets and repetitions that need to be done. This is then converted to specific information about the intensity and frequency of the exercises. When users follow a predefined workout plan, they can be informed about how close they are to achieving their personal goals such as weight loss, toning, or fitness level.

For high intensity workouts, where the user quickly switches between different activities, for example, exercising for 20 to 30 seconds, followed by 20 to 30 seconds of rest. The AI tracking device can automatically and reliably recognize each new exercise upon transition from one to the other.

Edge Processing. There are several advantages to having the data processing embedded in the sensor, rather than transmitting raw data to a central processor, perhaps in the cloud. First of all, it takes much more power to transmit large amounts of data, rather than a signal that has already used the data to compute a significant output message. Power can also be reduced if the local processor can aggregate data and just transmit it at fixed intervals rather than continuously. This is especially important for maximizing battery life. There’s also the question of privacy: home users, but especially professional athletes, do not want others to have access to their records. And for rapid motions you do not want the latency caused by sending data elsewhere rather than processing it right at the sensor.

Summing it Up

The development of ever more sophisticated MEMS sensors is enabling applications for wearables limited only by the imaginations of designers. Two examples of key advances at this point in time include the Bosch BMP384 gel-filled barometric pressure sensor, which is highly resistant to liquids and the BHI260AP, which includes an IMU and an edge processor for running artificial intelligence applications.

This article was written by Dr. Stefan Finkbeiner, CEO, Bosch Sensortec (Reutlingen, Germany). For more information, contact Constantin Schmauder at This email address is being protected from spambots. You need JavaScript enabled to view it.; or visit here .