A metamodel-based adaptive sampling approach for efficient failure region characterization of integrated circuits

Adaptive verification appears to be an e client solution to overcome the coverage problem and to accurately characterize the failure region of high dimensional spaces at integrated circuits' verification. Its main task is to gather more samples in the regions of interest based on the information learnt from previous samples. This helps engineers understand and interpret the behavior of the system under study with a reduced number of simulations/measurements compared to classical verification methods. To this end, we propose an adaptive sampling approach for the failure region characterization using the concept of metamodeling. Compared to other sampling methods for the failure region characterization, it has the advantage that it can detect and sample more in the near-failure region in the absence of a fail region. The concept has been applied on several synthetic test functions and on lab measurements of an analog integrated circuit. Results reveal that this adaptive sampling approach is very promising for failure region characterization.