Masala-CHAI: A Large-Scale SPICE Netlist Dataset for Analog Circuits by Harnessing AI

New York University - Tandon School of Engineering
Descriptive Alt Text
Dataset preparation flow starting from a PDF document consisting of Schematic and related text. Masala-CHAI extracts the relevant information and generates the SPICE netlist for the schematic.

Abstract

Masala-CHAI is a fully automated framework leveraging large language models (LLMs) to generate Simulation Programs with Integrated Circuit Emphasis (SPICE) netlists. It addresses a long-standing challenge in automating netlist generation for analog circuits within circuit design automation. Automating this workflow could accelerate the creation of finetuned LLMs for analog circuit design and verification. We identify key challenges in this automation and evaluate multi-modal capabilities of state-of-the-art LLMs, particularly GPT-4, to address them. We propose a three-step workflow to overcome limitations: labeling analog circuits, prompt tuning, and netlist verification. This approach creates an end-to-end SPICE netlist generator from circuit schematic images, tackling the long-standing hurdle of accurate netlist generation. Our framework demonstrates performance improvements, tested on approximately 7500 schematics of varying complexity across 10 books. As a part of the study, we have finetuned different LLM models with the generated dataset and reported their capability to generate SPICE netlists. We open-source this solution for community-driven development.

Descriptive Alt Text
Automating the conversion of analog circuit schematics into annotated schematics, enabling accurate SPICE netlist generation by Large Language Models.

Results

BibTeX

@misc{bhandari2025masalachailargescalespicenetlist,
      title={Masala-CHAI: A Large-Scale SPICE Netlist Dataset for Analog Circuits by Harnessing AI}, 
      author={Jitendra Bhandari and Vineet Bhat and Yuheng He and Hamed Rahmani and Siddharth Garg and Ramesh Karri},
      year={2025},
      eprint={2411.14299},
      archivePrefix={arXiv},
      primaryClass={cs.AR},
      url={https://arxiv.org/abs/2411.14299}, 
}