Simulated Scattering Parameter Dataset for a Randomly Generated Non-Parameterized Microstrip Bandpass Filter between 1 and 10 GHz
Description
The data set is comprised of two main components, a topology and a response, where each entry holds a topology and its corresponding response. The topology is a 32 x 32 binary image (and accompanying 32 x 32 binary array) where each pixel in the image represents the existence, or lack thereof, of conductor on the patch of the simulated model. The simulated model is created as a micro-strip filter with a 16 mm square patch placed in between 2 perpendicular 50 Ohm feed lines. The simulated model uses a 1.6 mm thick, 30 mm wide, square FR-4 substrate and a ground plane. The response of an entry contains the simulated electrical response of the simulated model, when the model’s square patch is updated to represent the entries topology, each pixel in the image represents a 0.5 mm square on the patch. The electrical responses recorded are the scattering parameters, more commonly known as the S parameters, the reference impedance and the frequency points at which the previous two parameters were calculated, 5001 points between 1 and 10 GHz. All electrical responses were created using the same simulation setup in CST Studio Suite. Each set of topologies, or each file contained within the two Topologies folders, are generated by iteratively sampling a topology distribution. A topology distribution refers to a set of parameters, parameter ranges and the probability distributions within these ranges. The choice of considered topology distributions, found in the Considered Topologies folder, are based off the electromagnetic response of an empty model, or a model with no topology added and perturbation theory. The 4 folders in this data set contain: MATLAB -> An example script to use the data set in MATLAB. Python-> An example script to use the data set in Python. Considered Topologies -> 800 entries used for data analysis containing 100 entries of each considered topology (X, Rotated X, Rotated Noisy X, Slash, Rotated Slash, Rotated Circle, H, and Y). Final Topologies -> 5000 entries created in response to the diversity analysis of the Considered Topologies, namely 3000 Rotated Noisy X and 2000 Rotated X. Scripts -> All MATLAB scripts and CST model required to recreate this dataset. Each file contained within the Considered and Final Topologies folders uses the file name structure <topologyIndex_topologyType-numberOfEntries.mat>, where topologyIndex is used to differentiate between 2 files containing the same topologyType. Both Considered and Final Topologies folders contain an Images folder; each topology is saved as an image to make it easy to find specific topology using the file name <topologyIndex_topologyType_imageIndex.png>. The images name is also stored as an additional field in each entry, in each topology set.
Files
Steps to reproduce
All code that creates the dataset can be found in the github repo attached below and the Scripts folder. CST Studio Suite 2023 and MATLAB R2024b were used. The file datasetSolver.mlx contains the code to generate the electrical responses for a given topology using CST Studio Suite. The code does the following in combination with datasetCreation.mlx: 1. Generate the binary 32 x 32 binary matrix that represents the topology to simulate, named the “topology”. 2. A MATLAB script, ShapeOperations/findShapes.m, takes the topology and generates the vertices for each shape within the grid. 3. The verticies are then used to update CST Models/microwaveFilter/Model/3D/grid_1.dxf. 4. MATLAB then uses an ActiveX connection to communicate with CST to do the following: a. Open the CST File b. Clear all previous results using CST Models/microwaveFilter/Model/3D/deleteResults.mcr. c. Clear the previous cutout, if any, using CST Models/microwaveFilter/Model/3D/updateGrid.dxf.mcr. This CST macro uses grid_1.dxf to cut the shapes out of the clear model. d. Start the simulation 5. Once the simulation is completed, retrieve the results from CST and save them to the relevant data file.
Institutions
- Stellenbosch University