Circuits-Informed Machine Learning Technique for Blind Open-Loop Digital Calibration of SAR ADC

This work presents a supervised machine-learning (ML) approach for blind digital calibration of SAR ADCs without requiring prior knowledge of errors. A 2-layer neural network learns the difference between outputs of a high-speed ADC and a low-speed, reference ADC when their sampling instants align and uses this knowledge to estimate and subtract errors in high-speed ADC at the back-end. The proposed ML-calibration improves SFDR of a 28nm, 12-bit, 84MHz ADC by >38dB while consuming 25.8fJ/conversion-step.