Hardware agnostic sensor networks
News Release from:
03 October 2012
System cost is a factor in deciding whether to use a sensor hub, points out Ian Chen, Executive Vice President of Sensor Platforms, a US startup which develops hardware agnostic sensor fusion software. Sensor Platforms' FreeMotion library of software algorithms and middleware supports all major mobile microprocessors and all sensors used in smartphones and tablets.
The company believes that a hardware agnostic approach allows developers the freedom to target different price/performance sensors – FreeMotion’s sensor component characteristics can be customised to work with a particular combination of sensors during system integration.
Chen described a tradeoff between power consumption and system cost for central versus distributed processing.
“This may be different based on the system design objective,” he says. “From a performance perspective, a dedicated sensor hub is usually much smaller than the application processor or main CPU of the system, so running sensor fusion there will consume less power. However, the application processor is the most cost efficient computing resource. Dedicating specific computing resources like an MCU or sensor hub for sensor fusion increases system cost.”
Chen also highlighted the fact that the application processor is not a perfect real-time computing resource. If there is a task with a higher priority running simultaneously, the application processor can pre-empt sensor fusion and cause it to miss samples or cycles. The real answer to where processing for sensor fusion should reside may lie somewhere in between the sensor hub and the application processor, he says, with a hybrid approach running sensor fusion in part on ‘smart sensors’ and in part in non-dedicated MCUs or on the application processor.
“In these scenarios, the most power consuming computation will be performed with efficient, dedicated hardware in a smart sensor, while the remaining computation would be distributed among other resources that already exist in the system, like the application processor or system management microcontroller,” he explains. “With intelligent architectural partitioning, this can offer the best of both worlds.”
Sensor Platforms’ software can be run on the application processor, on a sensor hub or spread between the two.
Also advocating hardware-agnostic software is Movea, whose MotionCore offering for mobile devices is software and firmware IP cores which are easily integrated into standard microprocessors and DSPs. As well as mobile devices, the company also has sensor fusion solutions for areas like fitness, such as the recently released Babolat Play & Connect tennis racquet which uses Movea’s MoveFit technology. The racquet uses MEMS sensors in the handle to detect stroke type, ball spin, power and other qualities.
You can read the rest of this article in the September issue of Electronic Specifier Design by clicking here.