A 14 bit 10MS/s TI SAR ADC with Neural Network Calibration

Abstract

In this paper, a four-channel 14bit 10M/S TI SAR ADC with 1.8V power supply voltage and 1V reference voltage is implemented on the Cadence platform. At the same time, a calibration method based on fully connected neural network is proposed. After the error analysis of the designed ADC, the random mismatch is introduced to the capacitance array(CDAC) and the PVT combination simulation is carried out in the center frequency range of ±1% to obtain multiple sets of error data as network training data and test data. The back propagation(BP) algorithm is used to train the neural network until the network converges, 11 sets of data are input for testing. The test results show that the average effective number of bits(ENOB) and spurious-free dynamic range(SFDR) are improved by 4.72bits and 34.17dBc respectively after calibration, and the network also has a calibration effect on input signals other than training frequency.