Estimation of noise power spectral densities from the Monte-Carlo simulated terminal currents in semiconductor devices

Abstract

A systematic technique for power spectral density estimation, called the correlogram method, which gives a statistically consistent spectral estimation with finite data sets, is applied to estimate the power spectral densities of noise currents in semiconductor devices with the Monte-Carlo simulated finite noise current data. Thermal noise and electron diffusion coefficients in an n-type homogeneous Si bar are estimated at various applied electric fields, and the estimated results are shown to be consistent and agree with the results obtained from other methods. Also the power spectral density of thermal noise in n- i- n structures operating near equilibrium is estimated, and the estimated results are shown to be consistent with the Nyquist theorem. Thus, we introduce a useful and powerful method of calculating accurately noise in submicron semiconductor devices, whose noise theories are not fully developed.