Quality Assurance of Advanced and Emerging Memory Technologies by Using Machine Learning – QUALMEM

Memory test is currently based on the use of March algorithms targeting Functional Fault Models. However, with shrinking technologies, these solutions will shortly become insufficient to achieve high coverage of new defect types, thus preventing quality assurance of memories after manufacturing. A solution to this problem is to investigate and adapt Cell-Aware test concepts successfully developed for logic circuits to advanced volatile (SRAM) and emerging non-volatile (MRAM) memories. The goal of QUALMEM is to develop test characterization models (called CA models) for gate-cells extracted from the memory description. These CA models will be enriched with layout information to allow complete coverage of realistic defects during test and diagnosis processes. Considering the significant number and the diversity of gate-cells for all considered memory technologies, Machine Learning techniques will be used for automating the generation process of the models in a time-efficient manner.