A Convolutional Neural Network Approach for Phased Array Calibration Using Power-Only Measurements

In this paper, a novel method based on a convolutional neural network is presented for receiver phased array online calibration using power-only measurements. Phase calibration coefficient for each RF path of a phased array receiver is estimated by combining a convolutional and multi-layer perceptron neural network in different signal to noise ratio (SNR) conditions. To validate the proposed method, a $1\times 8$ linear active phased array receiver is designed and simulated based on real characteristics. Compared to other conventional techniques such as the rotating element electric field vector (REV) method, our proposed method requires less number of power measurements and hence, is less time-consuming. Moreover, the proposed method has a significantly better performance in the presence of noisy measurements. Simulation results show that a mean-squared-error (MSE) of 6.5 is achieved at the signal to noise ratio of 20 dB for one iteration of the power measurement. These results prove that the proposed method can be adopted to the phased array online calibration, to have a faster calibration with higher accuracy in noisy conditions in comparison with conventional methods.