WiseEDA: LLMs in RF Circuit Design

Hangjiang Jin, Junchao Wang, Junjie Sheng, Yifan Wu, Jiayu Chen, Yaqi Wang, Jun Liu

Abstract: As the complexity of Radio Frequency Integrated Circuit (RFIC) design increases, the significance of Electronic Design Automation (EDA) becomes more pronounced. This paper proposes a circuit design methodology utilizing Large Language Models (LLMs), incorporating tools for topology selection and netlist optimization based on a particle swarm optimization algorithm. These tools are driven by LLMs, enabling engineers to describe their requirements in natural language. The LLM then selects suitable topologies and automatically configures relevant parameters for the optimizer, facilitating the automated netlist circuit design. Experimental results demonstrate that, after equipping the LLM with relevant knowledge through prompt engineering, it can optimize the values of capacitors, inductors, and other parameters in the band-pass filter netlist. This ensures that the filter’s S11 and S21 performance meet the specified requirements within a particular frequency band, thereby confirming the feasibility of employing LLMs in the automated circuit design process.