Design and analysis of internal calibration system for the Ku-band radar using learning algorithm with gradient-descent method

Abstract

This paper presents a novel approach to internal calibration of RF system. Ku-Band RF system with internal calibration path is designed to implement internal calibration process. Leakage signal path of the system is analyzed. From the analysis, robustness of system against leakage signal is demonstrated. The variation of system performance caused by thermal drift is compensated using learning algorithm with Gradient-Descent. System Offset Factor and Calibration Factor are newly defined for proposed internal calibration process. In learning algorithm, penalty function is applied to loss function with the analysis on local minima of loss function to improve stability of learning process. The results of internal calibration process of system are presented. Gain and phase difference in TX path are adjusted with the maximum variation smaller than 0.0477dB and $0.2481^\circ$, respectively. Also gain and phase difference in RX path are compensated with the variation less than 0.0132dB and $0.0722^\circ$, respectively.