While helping to improve motion-based accuracy in mobile and gaming devices, the advent of sensor fusion in the microelectromechanical systems (MEMS) space also will pave the way for significant new developments in centralized processing with substantial upside revenue, according to an IHS iSuppli MEMS & Sensors special report from information and analysis provider IHS.
The developments now unfolding in centralized processing are expected to work with current 9-axis sensor fusion—a technology that combines the use of 3-axis accelerometers, 3-axis gyroscopes and 3-axis electronic compasses to provide more accurate and sensitive motion detection. Revenue for motion sensors employed in a 9-axis combination will reach $850.97 million next year, up a notable 55 percent from $550.01 million at the end of 2011, on its way to some $1.35 billion by 2015. The impressive growth potential for sensor fusion is likely to redound to the developers of centralized processing as they tie their fortunes to fusion technology’s rising star.
The advances in centralized processing are meant to offload the heavy demand currently placed by smartphones, tablets and gaming consoles on the applications processor of those devices. Too often, apps processors require a great deal of power in order to run applications continuously in the background. Moreover, plenty of silicon chips end up getting duplicated across the various sensors implemented on a single mobile or gaming device to help detect a range of functions, including motion detection, ambient light, proximity, temperature and humidity. The result is an overloaded, power-hungry apps processor unable to function as quickly as envisioned or deliver maximum capability.
One approach for distributing the heavy load in apps processors is by the use of microcontrollers (MCU). Another—and potentially far more-reaching—advance will be the creation of an entirely new class of hardware dedicated to sensor processing.
The Case for MCUs
In the case of low-power MCUs, deploying them in combination with a cellphone’s apps processor will reduce power consumption, compared to simply relying on software algorithms that run on the apps processor. Furthermore, manufacturers can add their own code to the MCU to embed features that could help achieve product differentiation.
The main obstacle to using an MCU is cost: As an extra component, the MCU inevitably adds to the total bill of materials for the phone or any other device using the chip.
No mobile handsets to date have included a dedicated MCU for sensor signal processing, but some cellphone brands now are cooperating with MCU suppliers to develop solutions. The first phones with MCU sensor processing could appear as early as next year, IHS believes.
Among manufacturers, Texas Instruments was the first to implement motion sensor processing on its low-power MCU. Another company, San-Francisco-based Atmel Corp., also has announced its own cooperation deal earlier this year in May 2011 with InvenSense Inc. from Sunnyvale, Calif. The effort applies mostly to consumer applications requiring 6-axis sensor fusion, but Atmel is likely to develop its offer to sell its MCU into handsets for simple sensor fusion as well.
Other entities working with the dedicated MCU concept for handsets are Texas-based Freescale Semiconductor, which is promoting the concept of a “smart sensor hub” that combines its low-power MCU and accelerometer in the same package; and Italian-French manufacturer STMicroelectronics, which uses an internally produced MCU to run its own 9-axis sensor fusion algorithms as well as processing for ambient light and proximity sensors.
Across the seas, Japanese-based Rohm Semiconductor—owner of Kionix Inc. from New York—is also exploring the use of MCUs for sensor processing, IHS believes. Rohm already provides a range of ambient light and proximity sensors for handsets.
What a Dedicated Hardware Processor Could Achieve; MCU or Dedicated Hardware?
In the second solution to offload the heavy workload of the apps processor, suppliers of application processors envision the use of a new dedicated sensor-processing core, which can work while the rest of the apps processor is sleeping. Such a dedicated core will prove useful for functions that require continuous sensor operation during the day, as in the case with a device like a pedometer, or for instances involving activity monitoring and—eventually—context awareness.
With a dedicated core, sensor fusion would be extremely fast and efficient. For example, fewer clock cycles would be used to calculate coordinates. The hardware implementation also can provide many advantages, including much lower power consumption—especially important for mobile devices; as well as in eliminating duplicate silicon area for sensor signal processing. Manufacturers likewise need not develop a specific sensor solution algorithm, as they will be able to rely on the solution provided by the hardware in the apps processor.
The use of a dedicated sensor has its drawbacks. In particular, signifi cant investment from the apps processor developer will be required. And while the sensor fusion would be optimized for a specifi c range of sensors—typically those from market leaders—the same level of accuracy and resolution might not work out for every sensor from every other supplier.
Such potential challenges apparently are of little concern to chipmaker Qualcomm, which is believed to be introducing very soon a new system on chip (SoC) with dedicated hardware —an ARM 7 processor—for sensor fusion. This would fit right in with the brewing developments in the industry to reduce the workload on the apps processor.
Nonetheless, the move could be a bit too early for the market, IHS believes. While the accelerometer and compass side of the 9-axis fusion equation is mature, the technology governing 3-axis gyroscopes is still being improved. In particular, new lower-drift gyroscopes are expected to be rolled out in the next 18 months that likely will enhance the performance of the device. As a result, a hardware implementation of a dedicated sensor at this point won’t be able to take into account all the new gyroscopes arriving soon on the market.
Overall, the MCU approach makes more sense at present for offloading the apps processing workload. On the other hand, the use of a dedicated hardware could be optimal for power consumption and responsiveness after gyroscope technology matures and fully stabilizes.
Eventually, sensor fusion could have an additional impact on the motion sensor supply chain, and both MCU and processor suppliers could climb the value-chain ladder even higher, buying motion sensor elements or working with foundries to get the dies and then integrating them with their own integrated circuit chips.
Read More > Motion Sensors in Handsets and Tablets: It’s All About Fusion