A fast LMS-based digital background calibration technique for 16-bit SAR ADC with modified shuffling scheme

Abstract

-This paper introduces a modified shuffling scheme to achieve less overhead in ADC, lower power consumption for digital calibration and higher update frequency. Different-size unit capacitors are used in the shuffling, which reduces the number of switches and capacitors by almost half while remaining complexity of the digital circuit unchanged. Additionally, a novel LMS-based digital background calibration technique suitable for split-based ADC is proposed. Fast convergence and linear calibration are achieved with the improved gradient descent algorithm, which uses Momentum and Nesterov accelerated gradient (NAG). The convergence speed is found to be over 10 times faster than typical LMS algorithm and maintain calibration accuracy in MATLAB simulation of a 16-bit, 1 MS/s SAR ADC.