A Semiblind Digital-Domain Calibration of Pipelined A/D Converters via Convex Optimization

Abstract

This brief presents a semiblind and foreground calibration method for correcting linear and memoryless errors in pipelined analog-to-digital converters (ADCs). We formulate the calibration problem that finds optimal radices for maximizing signal-to-noise-distortion ratio as a linear fractional programming, a special type of convex optimization problem. The method is further extended to the case when the exact amplitude of a calibration signal is unknown. By utilizing the structure of the calibration signal, it is shown that the optimal radices can be obtained by solving the formulated calibration problem via bisection algorithm. Simulation results indicate that the proposed calibration method can correct linear errors in a hypothetical 14-bit 400 MS/s pipelined ADC using only $\approx 1600$ data samples.