SANTA CLARA, Calif. — PNI Sensor Corp. of Santa Rosa, Calif., demonstrated its new Sentral Sensor Fusion Hub coprocessor for the rest of the world this week at the STMicroelectronics Shaping the Future of MEMS and Sensors conference here.
The coprocessor is similar to Apple's recently announced M7 "motion coprocessor" chip for the new iPhone 5s. (See: China Is the Only Reason the iPhone 5c Matters.) Designed for any motion-tracking mobile device, the tiny Sentral coprocessor -- measuring just 1.5 by 1.5 by 0.5 millimeters -- constantly fuses the data streams from any accelerometer, gyroscope, and magnetometer (just like Apple's M7) to provide precise location, orientation, and heading information to any mobile device.
"We have taken the motion-processing algorithms we use with our own high-precision sensors and put them in a coprocessor chip that any mobile device can use with any brand of inertial sensor," Becky Oh, president and CEO of PNI, told us.
At the conference, PNI demonstrated the newly released Sentral coprocessor using ST's accelerometer, gryroscope, and magnetometers -- the same brand of sensors used in Apple's iPhone-4s (according to teardowns).
Diagram of PNI Sensor's Sentral motion processor, which it demonstrated the same day Apple unveiled its M7 motion processor for the iPhone 5s.
STMicroelectronics is in a dead heat with Bosch for the worldwide lead in the microelectromechanical systems (MEMS) chip market. The motion-processing fusion algorithms provided by PNI -- whose motion algorithms are used by everyone from NASA to the Nintendo Wii U GamePad -- up the ante in accuracy. Unfortunately, PNI's high-precision non-MEMS sensors are too big for slim-line smartphones. However, with PNI's Sentral coprocessor, any mobile device maker can get accurate motion-sensing algorithms in a tiny package.
PNI says on its website that its Sensor Fusion Hub coprocessor uses less than 1 percent of the power required to run similar motion-processing algorithms on a mobile device's application processor, because it uses a dedicated state machine, rather than firmware. The nine-axis auto-calibrating sensor fusion uses patented Kalman filter algorithms to provide more accurate readings than firmware.