ADAS High Bandwidth Imaging Implementation Strategies (Texas Instruments)
Abstract: MIPI CSI-2 specification, which was originally designed for mobile devices, is finding applications in automotive for Advanced Driver Assistance Systems (ADAS). ADAS imaging includes applications like parking assistance, collision avoidance, electronic mirrors, radar sensing, fusion and many more. These applications require high bandwidth data streams from multiple cameras and sensors on the vehicle. The upsurge in camera usage puts a strain for designers to explore methods that reduce pin counts, optimize board size, and retain high image resolution; features that are difficult to achieve with the traditional parallel camera interfaces. Additional challenges include; synchronizing all the cameras & sensors, meeting automotive class safety specifications, and enabling high image quality & wide dynamic range. This paper discusses strategies to meet the design goals using MIPI CSI-2 standard. The paper will use a DS90UB96x camera hub and TDAx processor as an example for discussion. The camera hub aggregates 4 independent video data streams via serial interfaces and outputs a collective signal in MIPI CSI-2 standard format to TDAx. TDAx receives the aggregated signal and efficiently separates individual camera streams using MIPI CSI-2 virtual channel information. DS90UB96x and TDAx also provide mechanism to synchronize the streams from all cameras. This setup provides the advantage of lower system EMI while achieving up to 6 Gbps output and color depth of 24 bits/pixel. The paper then discusses future expandability with improvements made in CSI-2 v2.0 specification, which allows connections of up to 32 cameras.
Shiou Mei Huang is an Automotive Hardware Applications Engineer at Texas Instruments (TI), based in Sugar Land, Texas, USA. Shiou Mei obtained her bachelor’s degree in electrical engineering from SUNY Stony Brook University before joining TI in 2012. She is currently responsible for product characterization and applications support in the automotive realm. One of her key roles is to ensure automotive infotainment and ADAS customers can release product to market in a timely fashion.
Mayank Mangla is an Imaging Architect at Automotive Processors Business Unit of Texas Instruments, based out of Dallas, Texas, USA. He is passionate about Digital Image Processing, being involved in the field since 2001. In the last 16 years se has done some pioneering work in camera algorithms and applications. He holds several patents and has authored numerous papers. His current focus area of research is ADAS (Advanced Driver Assist Systems) imaging, involving automotive applications like Electronic Mirrors, Park Assist Systems, Autonomous Driving etc.