Digital Adaptive Calibration of Multi-Step Analog to Digital Converters

Abstract

This paper reports a novel approach for calibration of multi-step A/D converters based on the steepest-descent estimation method. The calibration procedure is enhanced with dedicated embedded sensors, which register on-chip process parameter and temperature variations. Additionally, to guide the verification process with the information obtained through process monitoring, two efficient algorithms based on an expectation-maximization method and adjusted support vector machine classifier, respectively, are proposed. The algorithms are evaluated on a prototype 12 bits A/D converter fabricated in standard single poly, six metal 90 nm CMOS.

Behavior-Level Analysis of a Successive Stochastic Approximation Analog-to-Digital Conversion System for Multi-Channel Biomedical Data Acquisition

“…The validity and efficiency of complete-data-based methods cannot be guaranteed in such a case. Although a multiple-imputation method based on the expectation maximization was proposed in order to supplement the circuit calibration and to guide the verification process with the information obtained through the monitoring process, this method requires die-level process monitoring circuits [24]. The calibration techniques used in the present study do not require die-level process monitoring.…”

“…full-range single-tone signal input (frequency: 164.99 Hz). In order to obtain static characteristics with 99% confidence at 0.2-bit accuracy, the number of sampling points was set to 1024 . Figures12(a) and 12(b) show the differential nonlinearity (DNL) error in the output code calibrated by machine learning.…”