A Near-Field Super-Resolution Network for Accelerating Antenna Characterization

TL;DR

This paper tackles reducing data-collection time in near-field antenna measurements by introducing NFS-Net, a deep neural network that reconstructs fully sampled near-field data from undersampled measurements. The fully reconstructed NF data are then transformed to the far field via NF2FF, enabling accurate radiation pattern characterization with only about 11% of Nyquist-rate samples. A large, diverse simulated dataset and novel loss functions for magnitude and phase (including a periodic phase loss) underpin the training of a U-Net–based architecture. Experimental results on both simulated and measured antennas demonstrate high accuracy and generalizability, suggesting a practical pathway to accelerate NF antenna measurements across setups and frequencies.

Abstract

We present a deep neural network-enabled method to accelerate near-field (NF) antenna measurement. We develop a Near-field Super-resolution Network (NFS-Net) to reconstruct significantly undersampled near-field data as high-resolution data, which considerably reduces the number of sampling points required for NF measurement and thus improves measurement efficiency. The high-resolution near-field data reconstructed by the network is further processed by a near-field-to-far-field (NF2FF) transformation to obtain far-field antenna radiation patterns. Our experiments demonstrate that the NFS-Net exhibits both accuracy and generalizability in restoring high-resolution near-field data from low-resolution input. The NF measurement workflow that combines the NFS-Net and the NF2FF algorithm enables accurate radiation pattern characterization with only 11% of the Nyquist rate samples. Though the experiments in this study are conducted on a planar setup with a uniform grid, the proposed method can serve as a universal strategy to accelerate measurements under different setups and conditions.

Figures (12)

• Figure 1: The workflow of the proposed end-to-end fast NF antenna measurement. It starts with the undersampled measurement of $E_x$ and $E_y$ in the near-field of the AUT. The undersampled near-field data map is then fed into the NFS-Net to restore its fully-sampled counterpart, which is then further processed by the NF2FF algorithm to reconstruct the far-field radiation pattern.
• Figure 2: The overall architecture of NFS-Net. It consists of an encoder, a decoder, and skip connections between them to preserve high-resolution features and extract informative contents. Detailed dimensions and channels for the training process can also be found in numerical annotations.
• Figure 3: Comparison of the network's capability to conduct near-field super-resolution on data with a downsampling factor of 3. Undersampled NF and ground truth represent the low-resolution input and the fully-sampled, high resolution output of the network, with predicted NF showing the network's estimations based on low-resolution data. Antenna 1 is a corrugated horn antenna working at 8.7 GHz, while antenna 2 operates at 3 GHz as a Yagi antenna. The range of values is [0, 1] as shown on the color bar.
• Figure 4: Comparison of the simulated (fully-sampled) and reconstructed (3$\times$ downsampled) far-field patterns in both E and H plane. Antenna 1 is an X-band corrugated horn antenna, Antenna 2 is a Yagi antenna operating at 3 GHz, and Antenna 3 is a modified version of Antenna 2 tilted at an angle of $15 ^\circ$. None of these types of antennas is included in the training dataset.
• Figure 5: Training and validation loss decay of magnitude (a) and phase (b) for the network.