FaRBS: A new PoF based VLSI reliability prediction method

FaRBS (Failure Rate Based Simulation Program with Integrated Circuit Emphasis) is a new physics-of-failure based Very Large Scale Integration (VLSI) circuit reliability prediction method. With multiple failure mechanisms inh erently modeled and analyzed, FaRBS will make reliability engineers' life much easier by directly revealing each stress's acceleration effect at system level, thus helping reliability engineers carry out system level VLSI reliability predictions and derating/uprating analyses. Based on Physics-of-Failure (PoF) models of intrinsic failure mechanisms, FaRBS takes a straightforward top-down, bottom-up approach to reduce both modeling and prediction complexity. Detailed application breakdown from the system level reveals bottom-level device's operation profile. Devicelevel reliability characterization provides accurate operation-ba sed dynamic stress modeling by utilizing the physics-of-failure models. For each failure mechanism, the best-fit lifetime distribution is selected to provide the reliability prediction. With the bottom level device reliability prediction, the application-specific circuit and corresponding system reliability is further predicted by considering the system structure. To demonstrate FaRBS and verify its prediction capability, reliability predictions were performed on a microcontroller, DRAM and microprocessor. The predicted failure rates were compared with field data (from 2002 to 20 09) and demonstrate that the prediction agrees very well with the real failure rates.