Phased Array Calibration Using Residual Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional phased array calibration methods are inefficient and inaccurate due to the need for numerous measurements and the accumulation of errors, especially as the scale of integrated arrays increases.
Innovation Solution
A method using a residual neural network for fast automatic calibration of phased arrays, which involves setting a phase setting matrix, measuring amplitude and phase vectors, performing feature extraction, constructing a residual neural network, and using it to automatically estimate amplitude-phase errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If single-element calibration methods are used, then measurement accuracy is improved, but calibration efficiency deteriorates
Solution Approach 1:
The patent combines multiple array elements to perform parallel calibration measurements simultaneously. Instead of calibrating one element at a time, multiple elements are calibrated together by measuring their combined far-field signals, which dramatically improves calibration efficiency while maintaining accuracy through the use of neural network-based error separation algorithms.
Solution Approach 2:
The patent replaces traditional mechanical/iterative calibration procedures with a neural network-based automatic calibration system. The neural network model automatically estimates amplitude-phase errors from measured signals, eliminating the need for manual traversal of phase states and extreme value searching, thereby improving both efficiency and automation.
2Productivity
If multi-element calibration methods are used, then calibration efficiency is improved, but measurement accuracy deteriorates
Solution Approach 1:
The patent introduces a neural network model as an intermediary between the measured far-field signals and the amplitude-phase error estimation. This neural network intermediary processes the complex multi-element measurement data and separates individual element errors from collective measurements, enabling accurate calibration of multiple elements simultaneously without error accumulation.
Solution Approach 2:
The patent uses simulation software to generate large numbers of training datasets that copy realistic calibration scenarios. These synthetic datasets are used to train the neural network model, enabling it to learn and generalize error patterns from simulated measurements, which then improves the accuracy of actual calibration measurements.
3Adaptability or versatility
If array scale is increased, then array performance is improved, but calibration complexity increases
Solution Approach 1:
The patent replaces complex iterative calibration algorithms with a neural network-based automatic estimation system. The neural network directly maps measured signals to amplitude-phase errors without requiring iterative optimization or extreme value searching, significantly simplifying the calibration process for large-scale arrays while maintaining accuracy.
Solution Approach 2:
The patent implements self-service calibration where the system automatically performs calibration without manual intervention. The neural network model automatically processes measurements and generates correction parameters, eliminating the need for operators to manually traverse phase states or interpret complex measurement data, thereby reducing operational complexity for large arrays.
Data Source
AI summary
Disclosed is a method for fast automatic calibration of a phased array based on a residual neural network. A phase setting matrix is set and an amplitude and a phase of a array far-field complex signal are measured with a network analyzer to obtain an amplitude and phase vector of the array far-field complex signal. A real part, an imaginary part, and a magnitude of the far-field measured complex signal value are separated and normalized, and mapped to RGB three-channel image data. Datasets are automatically generated according to a preset amplitude-phase error range by a simulation software, the datasets are proportionally divided into a training set and a test set to be input into the residual neural network for training to obtain a calibration model. Measured data is input into the calibration model for automatic estimation of the amplitude-phase error of the phased array.