NEW YORK – As more and more MEMS sensors are showing up in mobile devices, the focus of MEMS design has begun shifting from discrete MEMS components to MEMS sensor data integration.
How raw data from multiple sensors is “fused” and “interpreted” makes a noticeable difference in a system’s power consumption and apps performance, according to Ian Chen, executive vice president, Sensor Platforms, Inc., a San Jose, Calif.-based start-up.
Keeping that premise as the company mission, Sensor Platforms is rolling out Monday (March 26th) a library of software algorithms and middleware designed, according to company claims, “to interpret users’ contexts and intents” by using data from multiple sensors in mobile devices.
Commonly found MEMS sensors in today’s smart phones and tablets include accelerometers, magnetometers, gyroscopes and barometers.
The report points out that one unanswered question confronting today’s MEMS component suppliers and system designers is: “Who will determine the sensor architecture, where the processing will reside and the motherboard-level sensor fusion architecture?”
While some companies may integrate all these sensors in a single monolithic device and wrap everything in smart components (although not an easy feat), Sensor Platforms has chosen another approach: offer sensor fusion in software that’s hardware agnostic.
The company says that its single code-base software can be used across platforms. It can be run in its entirety on an apps processor, on a sensor hub, or spread over the system.
That gives one clear advantage to Sensor Platforms’ software-based sensor fusion, said Chen: flexibility. “Our software allows system designers to pick and choose different supply sources for each sensor. The flexibility in sourcing is critical since these sensors come at different price and performance points.”
Another advantage of Sensor Platforms’ software lies in the conservation of sensor power, according to the company.
Throwing more sensors into a mobile device is one thing. But how to minimize the associated power consumption is another.
Chen noted, “Up to 10mW is added in power consumption when sensors are in use.” Sensor Platforms’ software library, called “FreeMotion Library,” comes with a proprietary algorithm that can “turn off power hungry sensors, like the gyroscope, and emulate its function with lower power sensors when user movements are slow.” That translates into “dropping sensor power consumption by 90 percent,” he added.
Then, there is the issue of reliability. Although not broadly advertised, some smart phones’ compass calibration can be off by 90 degrees, according to Chen. “All sensors require frequent calibration to maintain their data quality,” he noted. Sensor Platforms’ FreeMotion Library is built on an architecture that supports reliable sampling, and ongoing cross-sensor calibration to assure reliable sensor information – both for application developers and end users, the company said.
In sum, just designing a number of MEMS sensors into one’s system is hardly enough to improve a system’s power consumption, flexibility in system designs, sensor data or reliability and apps’ performance.
Tony Massimini, chief of technology at Semico Research, noted: “Now with all this data, how do the system designers fully utilize it? We may be just scratching the surface.” Further, he acknowledged, “The industry is at an early stage. System designers, many new to MEMS, need software development tools.”
The need for SDK is not limited to systems designers. It extends to apps developers, who are trying to leverage motion data – collected by sensors in a mobile device – in their new apps.
Chen said Sensor Platforms is rolling out the FreeMotion library’s API, so the data produced by its sensor calibration and sensor fusion provides apps developers robust data with better accuracy. Further, Chen said that “apps need information and context about the user, not just the user’s location or changes in his motion or direction.” The FreeMotion software development kit offers “a foundation to extend the type of information that applications can gain from sensor data,” according to Sensor Platforms.
Sensor Platforms’ key competitors are likely to be MEMS sensor component suppliers, who are developing their own sensor fusion software.
Earlier this year, Freescale Semiconductor, for example, introduced its own sensor fusion algorithms called Xtrinsic for electronic compass. Electronic compass applications combine magnetometer-provided headings with corrections from inertial sensors that compensate for stray magnetic fields. Freescale is offering its sensor fusion algorithms as a free download for its MEMS sensor users.
Meanwhile, STMicroelectronics last fall rolled out its own sensor fusion algorithms called iNEMO Engine Sensor Fusion Suite. According to ST, its iNEMO Engine can be combined with ST’s iNEMO Inertial Modules to create complete and customizable hardware/software multi-axis MEMS sensor solutions for enhanced motion and accurate heading recognition.
These MEMS sensor vendors use proprietary libraries to lock customers into their products, explained Sensor Platforms’ Chen, in contrast to his company’s hardware-independent sensor fusion software.
Semico’s Massimini, describing sensor fusion as “still at early stages,” expects “more innovation as it draws more attention.”
While the competitive landscape remains far from being defined, Massimini mentioned Movea as a potential competitor.
“Movea has been delivering solutions for several years while Sensor Platforms is a startup just hitting the market,” he said. Movea is offering motion-responsive software, firmware, and semiconductor IP for markets such as mobile and tablets, Interactive TV and sports and eHealth. Massimini noted, “Movea has developed a system that allows system developers to implement their algorithms using Movea IP blocks. Movea is working with a CAD vendor so the output of this tool can be used to design an ASIC.”
For mobile devices, sensor integration in both HW and SW is critical to meet the embedded requirements. IMHO, the application processor venders, instead of diversified sensor venders, need to offer the HW-agnostic fusion solution eventually.
The area of sensor fusion and the ability to filter and auto tune MEMs is just starting to take off. I can see opportunity in both the proprietary hardware/software as well as the more open source hardware/software approaches. The key is: performance/cost/feature tradeoffs. I would expect better performance/features in a proprietary system but at a cost. Likewise, the open ended system composed of varied vendor MEMs and a mix of software / hardware processing could yield reasonable cost/performance numbers. The overall trend towards more integration of algorithm processing with MEMs is a very encouraging one. I look forward to the continued growth and evolution that the marketplace will foster. I must confess that an integrated MEMs platform with auto-tuning/scaling and power savings is a very attractive offering.
Junko, there are many levels at which the sensor data can be 'fused' using software/algorithm-centric approaches as well as a combination of hardware and software/algorithms. Each have their advantages and access to being open and/or hardware neutral. The end application will decide what is better and the respective business model applies.
Take locating some one inside a building for example -there are many methods to triangulate and fine tune a moving person within a shopping mall (many companies in this space but none have it fine tuned to less than 10m accuracy yet). This involves fusing & processing data in real time from GPS, 3G/4G/LTE radio signals, WiFi, inertial sensors and BlueTooth, all of which is done with triangulation algorithms at the software-level, mostly hardware independent/agnostic.
In contrast, data from accelerometers and gyroscopes in the same package is best done at the hardware/software-level. Freescale's example cited in the article is one which allows non-Freescale micro's to access and process sensor data. But others like TI, ST, etc. may have proprietary algorithms that do compensations for noise in measured/sensed data and processing for linearity, etc., may choose to remain proprietary.
Sensors are Analog Elements used for sensing the real life signals, the entire instrumentation filed is focusing over the segment, but the sensors used in mobile devices are not yet fully the part of the instrumentation, but it has got equal amount of opportunities and possibilities instrumentation sensors and devices enjoys.
I’m writing on behalf of Hillcrest Labs, an expert in the kind of hardware-agnostic sensor fusion described in this article. As EE Times has covered, Hillcrest is known for selling its Freespace software and licensing its technology to companies such as LG Electronics, Roku,Sony,and others.
Today Hillcrest unveiled its new Freespace® MotionEngine™ for mobile and here’s an excerpt from the release: Hillcrest's core Freespace MotionEngine software, on which the new smartphone platform is based, was recently selected as a prestigious 2012 International CES Innovations Design and Engineering Awards Honoree. Benefits of Freespace include:
1) Sensor Agnostic Solutions: Hillcrest's MotionEngine software works with sensors from all major suppliers and Hillcrest can help companies qualify sensors based on specific performance and pricing needs. Hillcrest's sensor agnostic solutions enable lower costs by providing multiple supply choices to manufacturers and suppliers, while maintaining high performance standards.
2) Superior Sensor Performance: Hillcrest's proprietary static and dynamic calibration processes deliver the industry's highest motion processing performance, backed by proven mass-market deployments of major CE companies.
3) Minimal Power Consumption: With Hillcrest's software, sensors can be context-aware to know when they are needed or not, shutting on and off to preserve battery life.
4) Essential Patent Portfolio: Hillcrest's global intellectual property portfolio includes more than 60 issued patents, with more than 200 applications pending, and provides reassurance to customers bringing new motion-based products to market.
5) Flexible Implementation Options: With a modular software architecture and flexible implementation options for processor locations, sensor configurations, and operating system support, Hillcrest can support a wide range of applications and devices. More at www.hillcrestlabs.com
I suspect there are more vendors offering hardware-independent sensor fusion algorithms...but then, here's a question. What is their business model going to be? Is IP licensing enough to keep the business going?
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.