A Software Level Calibration Based on Bayesian Regression for a Successive Stochastic Approximation Analog-to-Digital Converter System

Recently, a novel low-power high-precision analog-to-digital converter (ADC) called the successive stochastic approximation ADC has been proposed which has two kinds of outputs from different modes, and which requires a software-level error correction method of combining them into a high-precision total output. From the practical viewpoint, we propose an error correction method based on the Bayesian regression with an incremental learning, in which additional data are successively selected according to the uncertainty of the corresponding predictive total output, and the uncertainty is approximately estimated by evaluating the upper bound of the standard deviations of the Bayesian predictive distributions of the outputs in each block of a partition of the all data set. Through numerical experiments, we verify the performance of the proposed method.