Microelectromechanical (MEMS) technology is a vital piece in the Internet of Things (IoT) puzzle, especially for mobile wearable technology (see “Wearing Your Technology”). It allows compact, inexpensive, accurate devices such as accelerometers to be standard fare in applications like smartphones.
I recently spoke to members of the MEMS Industry Group (MIG) about MEMS and IoT. MIG addresses the entire supply chain from foundries through end devices. The importance of MEMS is evidenced by the full-day conference track, Sensors & MEMS Technology, at this year’s Consumer Electronics Show (CES) in Las Vegas.
Wong: What roles do MEMS/sensors play in wearables and IoT devices?
Whalley: MEMS/sensors are the frontline “edge” devices collecting the raw data from the environment, such as pressure and temperature, or human body data, such as number of steps taken and heart rate. In wearable devices and IoT applications like smart homes, buildings, cities, and vehicles, they usually form a sensing cluster around the application processor, feeding it with every sensory change taking place. The individual data—or more commonly, combined or fused sensor data—is then processed using algorithms to make sense of this raw data that humans or machines can act upon.
Saxe: Sensors are the key to wearables and IoT, since they allow devices to “observe the world,” which in turn allows the devices to augment our own senses or autonomously make certain decisions for us. Of course, none of this works if the sensors are too costly, which is where MEMS comes in. By leveraging the incredible volume manufacturing capability of semiconductors, MEMS sensors can drive the cost of sensing down to the point that it is virtually free.
Wong: What is your view of sensor fusion for MEMS/sensor-based applications?
Whalley: Sensor fusion essentially makes the whole greater than the sum of the parts. Where multiple sensor data points and inputs can be combined, this provides a much richer picture of what is happening and what must be acted upon. The fusion can be data from the physical sensors on a device, and it can include information from “soft” sensors such as a calendar appointment, contact list, emails, flight information updates, etc.
It can also include data from Google Maps regarding, for example, traffic conditions, weather updates, pollen counts, etc. More information can be collected to make a more informed decision on what to act on. The key is not to confuse the user with too much information: Provide just the essentials, such as leave for the airport 30 minutes earlier than planned due to a traffic accident, or don’t run outside this morning because the air quality is in the red zone.
Saxe: Data is of no use unless it is turned into actionable information. Sensor fusion is key because it increases the quality of the information to the point that useful actions can be generated.
Consider the well-being case: 250,000 acceleration measurements taken over the course of a day are turned into information about how many steps you took and how much deep sleep you got, all through fusion. The next step is big data, which relates the steps to the quality of sleep and provides guidance on how to improve sleep.
Interesting things happen at the intersection of sensor data streams. Consider the case of a phone held up to an ear. The accelerometer can tell that the phone is in the right orientation, but doesn’t know if the phone is just being held up or is actually at the ear. The proximity detector knows if the phone is close to something, but doesn’t know if it is simply face down on the table or close to your head. By fusing the orientation information with the proximity information, we can be more confident that the phone is held to the ear, except maybe it is in a purse at about the right orientation and the proximity is to a book in the purse. If we could fuse that with an IR temp scan that told us the phone was close to something at 98.6°, then we could be even surer of its proximity.
Wong: What are the important new technologies for achieving lower power in consumer applications?
Saxe: There is no single answer. One high impact area is partitioning of algorithms between software and hardware. It is much easier to use software to develop algorithms, but hardware tends to require significantly less power. What’s required is an effective methodology or tool flow that achieves most of the flexibility and rapid development cycle of software, while harnessing a significant portion of the hardware’s power advantage.
Another high impact area is radio technology. BlueTooth Smart appears to have solved the low-data-rate, short-range challenge. There is no obvious winner yet for the medium-to-long range, low-data-rate challenge. This is important for home and building automation. BlueTooth Smart has the low power but not the range, and WiFi has the range but not the low power.
Wong: How will open-source initiatives, like MIG’s Accelerated Innovation Community (AIC), affect design-to-delivery of new MEMS/sensors-based applications?
Saxe: I think there are two phases to any open-source effort. When an algorithm is new, various parties will seek to gain a competitive advantage through their unique version of the algorithm. During this phase, expect to see only a basic algorithm submitted to the open-source repository, with the advanced algorithms remaining proprietary. Having the basic algorithm is still extremely valuable to applications developers, because it lets them sort out their core functionality using the basic algorithm in parallel with the development of the proprietary algorithms.
As the algorithm matures, a consensus forms as to what features and quality of results (QoR) are required. This reduces and eventually eliminates a proprietary algorithm’s competitive advantage. Subsequently, an “industrial quality” algorithm will be submitted to the open-source repository, which will become the de facto standard. Even proprietary implementations will support all of the features and QoR. Thus, applications will then be able to count on the features and QoR.
Whalley: The concept of Accelerated Innovation Community is to provide a forum that lets companies obtain basic sensor algorithms common to many designs, such as basic filters, step counters, 9-axis sensor fusion, etc. These baseline algorithms shouldn’t have to be created time and time again by each new company or application that comes along. The hope is that many companies, universities, research institutes, and individuals will contribute to AIC, and as users benefit from the algorithms, they will put back improved versions or derivatives of the algorithms and add new ones in the future, too. The goal is to speed product development times, lower costs, and foster innovation by allowing the company to focus on their true value-add and product differentiation.
The MEMS Industry Group will host a full-day conference track, Sensors & MEMS Technology, at 2015 CES. For more information, click here. MIG will also be on the show floor, January 6-9, 2015 during exhibition hours: Tech West, Sands Expo, Level 2, Booth 72032
