A Novel Fast Far-Field Phased Array Calibration Method Utilizing Deep Residual Neural Networks

Abstract

The calibration for a large phased array requires a significant amount of measurements using existing calibration methods. To accelerate the calibration process, this article proposes a novel fast far-field phased array calibration method utilizing deep residual neural networks. In the proposed method, a new feature extraction scheme (FES) is developed and applied to reconstruct the measured complex array signals in far-field into image data, which are then fed into the residual neural networks to train the calibration model. Specifically, the proposed one can train the calibration model based on datasets entirely generated by simulation, and the trained model can be directly applied across various array intervals, frequency bands, and calibration directions, given the same number of array elements. Consequently, the proposed algorithm excels in calibration efficiency as compared to conventional methods. For the verification, the proposed method is applied to the C-band, X-band, K-band, and Ka-band phased arrays that are produced based on our in-house integrated circuits (ICs). Measurement results obtained in the anechoic chamber validate the accuracy and robustness of the proposed scheme.