Statistical analysis of steady state leakage currents in nano-CMOS devices

Motivated by the problem of process variation in nano-scale CMOS, in this paper, we propose a multivariate statistical technique that uses the well known approach of principal component analysis (PCA), a technique extensively used in statistical modeling, to analyse the process variations. We use this approach to extract significant statistical information from the simulated data. We also propose a statistical model to characterize nano-scale CMOS device characteristics such as dynamic current I_dyn, gate leakage I_gate, and subthreshold leakage I_sub considering the effects of random variations in the process and design parameters such as gate oxide thickness T_ox, supply voltage V_dd and gate length L. These models can be used at higher level of circuit abstraction to study the design issues under process variation.