An artificial intelligence‐based 4‐to‐10‐bit variable resolution Flash ADC with 3.6 to 1.04 GS/s sampling rate

Abstract

AbstractThis paper presents an artificially intelligent Flash ADC with enhanced resolution from 4 to 10 bits. Unlike conventional approaches, this artificial intelligence (AI)‐based architecture avoids the use of many number of comparators in the Flash ADC when the ADC's resolution changes from 4 to 10 bits. This work initially gets the digital output of a 4‐bit existing Flash ADC as a training data set and then uses these 4‐bit output bits and sends to resolution enhancement logic (REL) block to vary its resolution without increasing the hardware complexities. After simulation, it is observed that the proposed ADC is of SNR of 24.13 dB for 4‐bit Flash ADC designed in SCL 180 nm CMOS technology and increases from 36.89 to 60.70 dB for 6 to 10‐bit resolution, respectively. The sampling frequency of the proposed architecture ranges from 3.6 to 1.04 GHz for a change in resolution from 4 to 10 bits. The FoM of 235 fJ/conv‐step in the training phase is obtained, and it varies from 56 to 20.3 fJ/conv‐step in the next phase of the testing and the prediction. The estimated area of the proposed 4‐to‐10‐bit variable resolution Flash ADC is μm2.