becklabash's picture
Upload README.md with huggingface_hub
b278902 verified
---
language: en
license: mit
pretty_name: Rectangular Patch Antenna Frequency Response Dataset
size_categories:
- 1K<n<10K
tags:
- patch-antenna
- antenna-design
- microwave
- rf
- electromagnetic-simulation
- inverse-design
splits:
- name: train
num_examples: 2051
---
# Rectangular Patch Antenna Frequency Response Dataset
Simulated S₁₁ frequency response curves for coaxial-fed rectangular patch antennas with varying dimensions and feed positions.
This dataset was released as part of [LaBash et al., "Improving Generative Inverse Design of Rectangular Patch Antennas with Test Time Optimization"](https://arxiv.org/abs/2505.18188).
## Dataset Description
Coaxial-fed rectangular patch antennas consist of a feed pin that passes through the ground plane to a metallic patch on a dielectric substrate. The design configuration of a single coaxial-fed rectangular patch antenna is parametrized by (*L*, *W*, *p*), where:
- *L* is the length of the patch in mm
- *W* is the width of the patch in mm
- *p* is the position of the feed point relative to the center of the patch along the length axis
<p align="center">
<img src="assets/patch_antenna_diagram.png" alt="Rectangular Patch Antenna Configuration - Top and Side Views" width="50%">
</p>
<p align="center">
<em>Figure 1: Configuration of a Rectangular Patch Antenna fed via coaxial line through the ground plane.</em>
</p>
### Design Parameter Ranges
- *L* = [7.5, 52.5] mm (patch length)
- *W*/*L* ratio = [0.8, 2] (width to length ratio)
- *p* = [-6, 0) mm (feed position), enforcing *p* = (-*L*/2, 0)
The designs were sampled at higher density at small *L* and with *p* close to the edge of the patch, then augmented using an algorithm designed to sample additional triplets (*L*, *W*, *p*) inside the convex hull of the existing dataset while enforcing uniformity.
### Simulation Details
The simulations were performed using [openEMS](https://openems.de/), an open-source electromagnetic field solver based on the Finite-Difference Time Domain (FDTD) method. Fixed substrate parameters were used:
- Dielectric constant εᵣ = 3.68
- Substrate thickness = 1.61 mm (aligned with OSH Park's 4-layer prototype service)
To calculate S₁₁ frequency response curves, each antenna was excited with a Gaussian pulse centered at f₀ = 5.5 GHz with a cutoff frequency fₖ = 4.5 GHz to cover the frequency range of interest, f∈[1GHz, 10GHz].
From the port data extracted through simulation, the complex amplitudes of the incident and reflected fields were obtained at *N* = 1000 regularly spaced frequencies. The reflection coefficient (S₁₁) was computed as the ratio of the reflected wave (u_ref) to the incident wave (u_inc), converted to decibels:
$$|S_{11}|_\text{dB}(f_i) = 20\log_{10}\left(\left|\frac{u_{\text{ref}}(f_i)}{u_{\text{inc}}(f_i)}\right|\right),\quad i=1,\dots,N$$
<p align="center">
<img src="assets/s11_example.png" alt="Example S11 Frequency Response" width="50%">
</p>
<p align="center">
<em>Figure 2: Example S11 vs. Frequency plot</em>
</p>
## Dataset Structure
Each sample in the dataset contains:
- **Design Parameters**:
- `length`: Patch antenna length in mm
- `width`: Patch antenna width in mm
- `feed_y`: Feed point position in mm relative to the center of the patch, along the length axis
- **Frequency Response**:
- `frequencies`: Array of 1000 frequency points per sample (Hz), ranging from 1 GHz to 10 GHz
- `s11`: Array of 1000 S11 values (dB) corresponding to each frequency point
- **Metadata**:
- `id`: Unique identifier for each sample
## Citation
If you use this dataset in your research, please cite:
```
@misc{labash2025improvinggenerativeinversedesign,
title={Improving Generative Inverse Design of Rectangular Patch Antennas with Test Time Optimization},
author={Beck LaBash and Shahriar Khushrushahi and Fabian Ruehle},
year={2025},
eprint={2505.18188},
archivePrefix={arXiv},
primaryClass={eess.SP},
url={https://arxiv.org/abs/2505.18188},
}
```