Skip to Main Content
We present a generic data-driven nonparametric analyzer (GDNA) to estimate the impact of process variations on device properties and circuit functionalities in nanometer technologies. The mathematical framework of GDNA uses a kernel estimator that eliminates the need for a priori assumption of the nature of variation (i.e., no a priori choice is required for the density of a random variable). Furthermore, a generic tail probability estimator is developed that uses the kernel estimator to predict low occurrence probabilities using a small set of observed samples. Verifications using statistical simulations show that GDNA can reliably predict variability in device/circuit properties and can hence facilitate technology development and circuit design under process variation.