CNN-Based Auto Fully-Digital Compensation Method for User Equipment with IQ Imbalance

Abstract

IQ imbalance (IQI) of radio-frequency (RF) Chains is one of the key factors that limits the performance of wireless communications systems, especially in high order modulation modes. Traditional methods usually estimate the IQI parameters with dedicated feedback circuits and compensate the distortion of RF Chains with extra compensation circuits. Moreover, they do not consider the random carrier phase shift between transceivers, which increases the difficulty of estimating IQI parameters. In the paper, we propose a CNN-based auto fully-digital calibration technique with no dedicated feedback and compensation circuits to compensate for IQI and the random phase shift. We train the CNN using the pilot signals to obtain the compensation matrix. Since only the pilot signal is adopted, the algorithm can work with current wireless communications systems automatically. Compared to user equipment (UE) with ideal RF chains, the simulation results show that our algorithm can help UEs with IQI and the random phase shift achieve similar performance even in 1024-QAM modulation mode.