SMPS System using Neural Network Controller with Distributed Pass Gates Optimized for Deep Neural Network Loads

Switched Mode Power Supplies (SMPS) and Low Drop-Out Regulators (LDOs) are extensively used in many Power-Management (PM) systems providing large currents to dynamic loads with excellent power efficiencies. Especially in recent years, these blocks are used to provide a dedicated power supply to different consumer electronics and applications such as Central Processing Units (CPU), Application processors (AP), and Communication Processors (CP) of mobile phones, tablets, or computers. These PM systems have digital and/or mixed-signal controllers on CMOS Integrated-Circuits (IC). Separately, there is also enormous interest to Artificial Intelligence and Machine Learning (AIML) devices and applications. Hence, very complex Deep Neural Network (DNN) and Convolutional Neural Network (CNN) systems requiring great computational power are increasingly implemented. In this paper, a novel idea is proposed to develop intelligent SMPS and LDO blocks being capable of managing drastic load changes and activity due to AIML algorithms and highly complex DNN structures.