An efficient High-Volume Production Performance Screening using On-Chip Ring Oscillators

Performance screening is an essential test for modern automotive microcontrollers (MCUs). Such a test determines the maximum achievable frequency F
MAX
of the MCU. On-chip ring oscillators (ROs) are used as indirect monitors for performance screening. The performance is determined by using the RO frequencies. This work uses various ROs to train a classification-based machine learning model that does a pass/fail classification based on a preset threshold value. An accuracy of over 97 % is achieved. The proposed performance screening flow selects the most appropriate subset of ROs and extracts sharp test limits that are directly applicable to the productive test flow.