On Energy-Reliability Tradeoff in Analog-to-Digital Converters with Imperfect Comparators

The paper investigates the tradeoff between energy-per-conversion and comparator reliability for Flash, first order SigmaDelta, and Beta expansion analog-to-digital converter architectures. Upper bounds on the minimum number of index bits required to achieve a prescribed distortion are derived as a function of comparator reliability. These bounds are then combined with a comparator energy-per-comparison-reliability relation to provide upper bounds on the energy-per-conversion for each architecture. The upper bounds are found to be monotonically increasing in reliability, suggesting that it is better from an energy-per-conversion perspective to use relatively unreliable comparators.