Adaptive Algorithms for Measuring the Parameters of Low-Frequency Noise of Semiconductor Devices Under the Conditions of Mass Control

Abstract

We solve the problem of increasing the reliability and efficiency of quality control for semiconductor devices and present the analysis of the conditions of measuring the power spectral density of low-frequency noise in semiconductor devices with spectra of the form G(f) ~ f–γ (γ is the parameter of shape of the spectrum) under the conditions of mass quality control. The measurement error of the power spectral density under given conditions of measurements strongly depends on the value of the parameter of form of the spectrum. We propose adaptive algorithms aimed at measuring the parameters of low-frequency noise in the cases of a given ultimate measurement error of power spectral density and a given duration of a single measurement. The proposed algorithms include the preliminary evaluation of the parameter of shape of the spectrum with subsequent measurements of the power spectral density of noise for the optimal bandwidth of the filter. The optimal bandwidth of the filter is established according to the results of preliminary evaluation of the parameter of shape of the spectrum. In both cases, we obtain estimates of the gain in a sense of the mean value over the set (ensemble) of controlled products. We also discussed the possibility of adaptive or cognitive adjustment of the parameters of measuring system in the control process based on the results of evaluation of sample means in the training sample.