| Proj No. | A3011-251 |
| Title | Deep Learning Approach for Microwave Active Circuits |
| Summary | Low-cost and accurate models of microwave transistors are indispensable in simulation of active circuits. This, in turn, is essential for efficient circuit characterization and design. Design procedures generally require behavioral models of the transistors, which adequately represent their large- and small-signal characteristics over a range of biasing conditions. Standard transistor models for microwave and RF applications usually employ physics-based equations. However, such models come short in certain aspects, e.g., reliable capturing of the electrical characteristics. Furthermore, the development of modelling equations for new physical phenomena requires considerable expertise. Finally, parameter extraction for equation-based models is challenging and difficult to automate1. An alternative for equation-based models are lookup table (LUT)-based methods. However, these techniques require long SPICE simulation time, as well as suffer from convergence issues for large-scale designs. In addition, LUT-based models lack the control parameters that can be used to manipulate the output characteristics of the model. Over the recent years, the role of Artificial Intelligence (AI)-based techniques has been continuously growing in the development of efficient numerical procedures for RF and microwave engineering, including data-driven surrogate modeling methods. AI-based modeling has the potential to tackle the limitations conventional approaches. This project is to attempt on developing the ANN model on the known transistor data. |
| Supervisor | A/P Arokiaswami Alphones (Loc:S2 > S2 B2A > S2 B2A 29, Ext: +65 67904486) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Communication Research II (Loc: S2-B3c-26) |
| Single/Group: | Single |
| Area: | Wireless and Communications Engineering |
| ISP/RI/SMP/SCP?: |