Experimental Multiport-Network Parameter Estimation for a Dynamic Metasurface Antenna

TL;DR

This paper presents an experimental method to accurately estimate the multiport network parameters of a dynamic metasurface antenna at 19 GHz, improving modeling accuracy over traditional simulation-based approaches.

Contribution

The authors develop an experimental parameter estimation technique for DMA forward models using multiport network theory, addressing challenges of non-linearity and measurement limitations.

Findings

Proxy MNT model achieves 40.3 dB accuracy in reflected field prediction.

Proxy MNT model achieves 37.7 dB accuracy in radiated field prediction.

Including auxiliary calibration feeds improves model accuracy for limited DMA configurations.

Abstract

Most use cases of reconfigurable antennas require an accurate forward model mapping configuration to radiated field (and reflections at feeds). Emerging dynamic metasurface antennas (DMAs) confront the conventional approach of extracting such a model from a numerical simulation with multiple challenges. First, the cost of accurately simulating an intricate and electrically large DMA architecture might be prohibitive. Second, the model-reality mismatch due to fabrication inaccuracies might be substantial, especially at higher frequencies and for DMA architectures leveraging strong inter-element mutual coupling (MC) to maximize their tunability. These considerations motivate an experimental parameter estimation for DMA forward models. The main challenge lies in the forward model's non-linearity due to inter-element MC. Multiport network theory (MNT) can accurately capture MC but the MC parameters cannot be measured directly. In this article, we demonstrate the experimental estimation of a high-accuracy proxy MNT model for a 19-GHz DMA with 7 feeds and 96 elements, where all feeds and elements are strongly coupled via a chaotic cavity. For a given DMA configuration and excitation, our proxy MNT model predicts the reflected field at the feeds and the radiated field with accuracies of 40.3 dB and 37.7 dB, respectively. A simpler, MC-unaware benchmark model only achieves 2.6 dB and 3.3 dB, respectively. We systematically examine the influence of the number of feeds and measured DMA configurations on the model accuracy, motivating the inclusion of "auxiliary calibration feeds" to facilitate the parameter estimation when the intended DMA operation is limited to a single feed. Finally, we measure DMA configurations optimized based on our proxy MNT model.