SoC architecture is optimised for Advanced Driver Assistance Systems

October 13, 2014 // By Christoph Hammerschmidt
The challenge in camera-based Advanced Driver Assistance Systems (ADAS) is to standardise the various output signals and turn them into a distortion-free image. Semiconductor design company Dream Chip has developed a uniform SoC architecture that will reduce the complexity of such devices and provides for a functional unit for image signal pre-processing.

According to Dream Chip, experience shows that purely software-based approaches do not meet the computing power requirements of certain application-specific image transformations. Instead, raw image signal pre-processing in the camera translates into a decisive difference in terms of scalability and cost. Based on this perception, Dream Chip has developed a technology that assumes the task of pre-processing these data and that enables designers to integrate multiple different cameras on a single SoC.

Dream Chip's architecture offers the possibility to chose between pure hardware and pure software implementations as well as any combination thereof, according to the requirements for resource efficiency, power efficiency and flexibility. It will fully exploit the potential of the recently introduced European 14/28 mm FDSOI (Fully Depleted Silicon on Insulator) process in terms of cost and complexity reduction. Furthermore, the architecture enables implementing diverse camera applications for ADAS on a single SoC - for instance a 360° top view camera, a stereo front view camera and digital rear view mirror applications with up to four cameras.

Dream Chip, a former Infineon spin-off based in Hannover, Germany, plans to introduce its architecture at the IZB automotive supplier conference and trade fair that takes place this week in Wolfsburg.

Related Links and articles:

New DSP IP core targets mobile, DTV, automotive and computer vision

Hi-resolution image recognition for driver assistance systems

Intelligent imaging for automotive safety