Toward a Faster Screening of Faulty Digital Chips via Current-Bound Estimation Based on Device Size and Threshold Voltage

摘要

Observations of peak and average currents are important for designed circuits, as faulty circuits have abnormal peaks and average currents. Using current bounds to detect faulty chips is a comparatively innovative idea, and many advanced schemes without them use it as a component in statistical outlier analysis. However, these previous research works have focused on the discussion of the testing impact without a proposed method to define reference current bounds to find faulty chips. A software framework is proposed to synthesize high-performance, power-performance optimized, noise-immune, and low-power circuits with current-bound estimations for testing. This framework offers a rapid methodology to quickly screen potential faulty chips by using the peak and average current bounds for different purposed circuits. The proposed estimation technique generates suitable reference current bounds from transistor threshold voltage and size adjustments. The SPICE-level simulation leads to the most accurate estimations. However, such simulations are not feasible for a large digital circuit. Hence, this work proposes constructing a feasible gate-level software framework for large digital circuits that will serve all of simulation purposes. In comparison with transistor-level Nanosim simulations, the proposed gate-level simulation framework has a margin of error of less than 2% in the peak current, and the computation time is 334 times faster.