Performance of Triple Mixed-ADC Large-Scale MIMO Detector Using Deep Learning

Large-scale multiple-input multiple-output (MIMO) transmission has been widely proposed for the current and future communication systems since this transmission technique opens up new options to expand the system capacity and quality of services for the radio networks. One-bit analog-to-digital converters (ADCs) and mixed-resolution ADCs are emerging techniques to reduce the power consumption and hardware cost when an enormous number of receive antennas are employed in the system. In this paper, we present the performance of the deep-learning-based detection of the triple mixed-ADC system, where the receiver side splits the set of receive antennas into three subsets and utilizes the extremely low-resolution, low-resolution, and high-resolution ADCs in those three subsets. We present the triple mixed-ADC complex channel and show the derivation to convert such a complex channel model to the equivalent real binary channel model to leverage the previous Deep Learning framework for detecting the large-scale MIMO signal. The model for the triple mixed-ADC is trained and tested successfully, and the experiment results show the advantages of the proposed triple mixed-ADC system over the single ADC and dual mixed ADC systems.