Machine Learning Assisted Direct Synthesis of Passive Networks
Passive matching networks are used extensively in RF/mm-Wave circuits and systems to ensure connections between devices/circuits with desired performance properties. such as maximum power transfer, load-line matching. noise matching, and passive voltage/current scaling. Although there are many passive network topologies to perform impedance matching, one commonality is that the process of designing, verifying, and optimizing these networks requires extensive use of EM simulations, such as Ansys HFSS. As a result, the design of these networks is oftentimes very tedious, time-consuming, and requires extensive design experience and computation resources. Only after constructing and simulating the EM structure can the designer then extract circuit performance metrics such as S-parameters, Z-parameters, loss, bandwidth, and examine the whole passive network performance. As the complexity of the passive network structure grows, simulation times may easily reach in the excess of several hours. This prevents the designer from performing optimization because this design and optimization process usually requires extensive iterations, further increasing the simulation time needed.
The objective of this project is to develop a machine learning-based solution for automating the design of transformers to achieve optimal power matching in electrical systems. The solution will be based on a mathematical model that relates the target load impedance to the passive network parameters and the transformer parameters to the transformer geometry.
The following steps will be taken to achieve the project objectives:
- Development of Equations: The first step is to develop the mathematical equations between the target load impedance and the passive network parameters, as well as the equations between the transformer parameters and the transformer geometry (e.g., having a closed form equation for the transformer’s inductor as a function of the available metal stack dimensions.). This may involve using curve fitting techniques to accurately model the equations.
- Verification of Model: The next step is to verify the accuracy of the model by comparing the results to real-world data. This will involve testing the model on a range of different loads and evaluating its performance.
- Program Development: Once the model has been verified, a program will be developed to transfer the user's desired target load impedance into an optimum transformer design in terms of bandwidth and passive efficiency. The program will be user-friendly and allow for easy customization and optimization.
- Expansion to Other Passive Architectures: The final step is to map the idea of automation design to other passive architectures, e.g., couple Line based balun design.
In this project, the student will:
- Develop the theoretical equations for the transformer.
- Get to know and simulate RF/mm-Wave passive structures using ADS.
- Know how to ideally model the transformers.
- Get to know and use EM simulators.
- Get to use MATLAB to write custom code to optimize the parameters to reach the optimum solution.
- Analog Integrated Circuits
- Electromagnetic Simulation Experience
- 10% Literature Survey
- 30% Layout and Electromagnetic Simulations
- 60% Develop Theoretical analysis
Last update: 13.02.2023