A memory volume diagnostics methodology to facilitate production yield learning with embedded memories

As the SOC area dedicated to memory is increasing, and the geometry is shrinking in advanced process technology, the yield fallout linked to memory is becoming more important. Traditional methods are proving either ineffective or very costly for production usage. A new methodology is required to extract failure data for analysis to help towards speedy yield learning and drive yield improvement. This paper describes a production based volume diagnostics methodology using a low cost ATE to obtain precise information on memory failures throughout the chip, and gain insight into systematic failures. This yield learning also provides valuable information to help improve for future designs.