Processing power targets sensor fusion

“Some people prefer to go for a dedicated sensor hub or microcontroller that takes care of everything,” Esfandyari says. “They are hoping they can put the application processor into sleep mode so they can reduce the power consumption… plus, if you have a dedicated sensor hub you don’t have to worry about the architecture of the application processor,” he adds.

As for the future, Esfandyari said that he sees a move away from the sensor hub towards the application processor as more sensors are required for further degrees of freedom and thus, more processing is required. He also explained that ST is not working to combine the sensor and the microcontroller in the same package.

“If you put the sensors and the microcontroller in the same package it gets very big, so you will lose a certain amount of flexibility in terms of where you can put the chip on the board,” he says.

There are also constraints for where a magnetic sensor can be placed on the board because of magnetic distortion. Gyroscopes too are sensitive to mechanical stress, so putting these components in one package would limit where it could be placed..

Stephane Gervais-Doucent, Freescale’s Global Marketing Director for sensors, agrees that the most power efficient approach is currently to use a sensor hub.

“Sensors require constant scanning and low to medium MIPS… power consumption should be minimised, hence, distributed processing is the key technology for sensor fusion,” he says. “This could be a sensor hub or it could be a small dedicated processor embedded in the application processor to avoid using one of the main CPU cores.”

Freescale recently launched its 12-axis Xtrinsic sensor fusion platform that has been demonstrated and certified for Windows 8. The platform, intended to add sensing capability to Windows 8 tablets and smart devices, incorporates Freescale’s accelerometer, gyroscope, magnetometer, barometer, temperature sensor and ambient light sensor. It uses a Freescale ColdFire+ MCU to process the sensor data.

Asked why sensor fusion is necessary, Gervais-Doucent gave a basic example of a compass application which combines accelerometer and magnetometer data. The software for magnetometer calibration and tilt compensation are what make a compass application useable in a smartphone. More complex sensor fusion software enables the combination of more sensors to yield elaborate gesture recognition. Unlike ST, Freescale is already combining these sensors and processing resources into the same package.

“For instance, we are able to implement in a tiny 3 by 3 mm package, either a 3-axis accelerometer with a 32-bit MCU or a 3-axis accelerometer with a 3-axis magnetometer,” Gervais-Doucent said.

You can read the rest of this article in the September issue of Electronic Specifier Design by clicking here.