Creating Reliable SoCs For Safe ADAS Applications

Every major automaker is in the process of bringing out autonomous vehicles with ADAS (advanced driver assistance systems). In addition to processors and embedded software, ADAS requires a variety of sensors – ultrasonic, camera, RADAR (radio detection and ranging), LIDAR (light detection and ranging), GPS and IR (infrared) – that are used to recognize signs, people, animals, other vehicles, road barriers, and more. The need to recognize data from the sensors built into the vehicle (in-vehicle), from other vehicles (vehicle-to-vehicle) and between a vehicle and transportation infrastructure (vehicle-to-infrastructure) under all lighting conditions is paramount, making the automotive ADAS one of the most safety-critical consumer applications.

The intelligence needed to parse data from a variety of sensors and almost immediately analyze and decide the course of action is typically designed into a system on chip (SoC) – the brains of the system! The chip needs to have the performance and capacity to operate on the variety and volume of data, and in the time needed. Any on-chip failure cannot be tolerated. To ensure the highest order of safety, very strict requirements are imposed on the design, verification and validation of such SoCs, both stand-alone and in the context of the system. These chips and the system they go into are required to operate not only with integrity and reliability for a period of 7 to 10 years, but also under a wide range of operating modes and conditions.

Capacity and performance requirements needed to handle the large volume of data from a variety of sources call for employing large SoCs, designed using advanced process technologies, which also use lower operating voltages. Since voltage drop directly impacts performance, ensuring power integrity for all operating modes and conditions is a critical requirement. Thermal conditions impact device timing, electromigration (EM) – failure over time – mechanical stress and electro-static discharge (ESD). Electromagnetic interference (EMI) is yet another reliability issue, and the SoCs are required to achieve target levels of electromagnetic compliance (EMC). All of the above reliability issues need to be identified and addressed to ensure reliable operation and safety.

Safety is best illustrated using an example. Consider a bicyclist riding along in the bicycle lane when all of a sudden the door of a parked car is opened directly in her path. To avoid crashing into the door, the cyclist veers slightly into the path of a car that is about to overtake her on the left. The ADAS application in the traveling car has to, in real-time, recognize the cyclist veering into its path, process and analyze data from all the related sensors and either issue a warning to the driver, or directly instruct the vehicle to apply the required corrective action that will avoid taking out the cyclist. This must be done in real-time and under any lighting conditions – a truly daunting challenge, both for the chip and system design.

Requirements such as high-quality image recognition and fast time-to-decision often call for the direct implementation of protocols like convolutional neural networking (CNN) into the chips. Even if you are the IC designer, with reliability and safety at the top of the list, bringing to market a chip that drives an autonomous vehicle system takes much more than verifying and validating the chip by itself. It needs to be verified and validated in the context of the entire system.

Various estimates state that billions of miles of road testing will be necessary to exercise the ADAS and autonomous driving system, including the SoC, to ensure reliability and safety. It is not only an expensive and time consuming proposition, it covers only a finite number of scenarios – practically speaking, a hundred or more. The key question is, what is the risk that an OEM is taking on?

On the other hand, simulation using precise modeling allows virtual testing across thousands of scenarios by quickly varying design parameters. Simulation covers more scenarios, saves time and money, reduces risk, improves product reliability, quality and safety. It also accelerates verification and validation of SoCs along with the complete ADAS and autonomous systems in six different areas.

Also, make sure you check out the ANSYS webinar on design methodologies to address finFET challenges related to power integrity and its impact on timing, thermal issues and reliability.