AaLLM: An End-to-End Analog Circuit Design Framework from Topology Generation to Sizing Using Large Language Models
AaLLM is an open-source end-to-end multi-agent LLM workflow that takes user specifications as input and outputs a netlist, encompassing both topology generation and circuit sizing. It automates creation of a knowledge base from research papers and textbooks and implements a RAG model.
Analog circuit design is time-consuming and relies on expert intuition. Existing LLM-based approaches are fragmented, focusing only on sizing or topology generation, require manual technical knowledge, are prone to hallucinations, and cannot create innovative topologies. AaLLM addresses these by providing an end-to-end multi-agent LLM workflow that generates both topology and sizing from user specs, using an automated knowledge base and RAG.
AaLLM uses a multi-agent LLM architecture with RAG over an automatically constructed knowledge base from papers and textbooks. This reduces manual data collection and aims to mitigate hallucinations during circuit sizing. The framework outputs netlists directly from specifications, covering topology generation and sizing in one workflow.
This research signals a shift toward automating analog circuit design, a domain traditionally requiring deep expertise. Open-source release may accelerate adoption in EDA and semiconductor design, potentially reducing design cycles and enabling non-experts to explore circuit design.
Potential to reduce analog design time and cost, lower expertise barriers, and improve design exploration. Could be valuable for semiconductor companies and EDA vendors seeking AI-assisted design tools.
Next observable signals include publication of benchmark results against conventional tools, community adoption of the open-source code, and integration with commercial EDA flows. Further work may address innovative topology generation and multi-spec trade-offs.