【Abstract】

  In recent years, the demand of AI-inside Automotive electronics (abbreviated as AA-chip in the sequel) has increased rapidly due to numerous emerging applications such as autonomous driving and vehicle-to- vehicle communication enhanced smart transportation system. The ICs in a smart automobile often need to support not only high performance computing, high-bandwidth communication, AI reasoning, but also high security, high functional safety, and long lifetime. All these requirements combined has led to many new technical challenges that mandate new solutions. For example, from the aspect of the functional safety, an AA- chip should be able to operate normally under extremely harsh operating environments (e.g., -40°C–150°C temperature range and radiation). Besides, they need to be able to cope with significant disturbances inside the chip induced by the AI-based high-performance computing such as significant temperature and/or IR-drop. Furthermore, they need to satisfy the stringent functional safety requirement of almost zero failure rate throughout a lifetime of 10 to 15 years. For comparison, a consumer electronics often require a Defective Level of a few hundred DPPMs (Defective Parts Per Million), but for AA-chips, we need to further reduce it down to almost zero DPPM. This is a challenging zero-in process that requires a much more comprehensive test methodology.