AnalogCoder-Pro: Unifying Analog Circuit Generation and Optimization via Multi-modal LLMs

2025-08-04

AI Summary

Analog front-end (AFE) design has long relied on expert intuition and iterative simulation, lacking fully automated methods capable of jointly optimizing topology and device sizing while generating performance-customizable, complete netlists. Method: This paper proposes the first generative, multimodal large language model–driven end-to-end joint optimization framework. It integrates functional specification text and waveform images as multimodal inputs, combines circuit netlist semantic parsing with simulation-guided optimization loops, incorporates rejection sampling–based fine-tuning to enhance generation quality, and establishes an automated parameter extraction and search-space modeling pipeline. Contribution/Results: Experiments demonstrate significant improvements in design success rate and circuit performance. The framework achieves, for the first time, fully automatic, performance-customizable schematic generation of high-performance analog front-ends.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Despite advances in analog design automation, analog front-end design still heavily depends on expert intuition and iterative simulations, underscoring critical gaps in fully automated optimization for performance-critical applications. Recently, the rapid development of Large Language Models (LLMs) has brought new promise to analog design automation. However, existing work remains in its early stages, and holistic joint optimization for practical end-to-end solutions remains largely unexplored. We propose AnalogCoder-Pro, a unified multimodal LLM-based framework that integrates generative capabilities and optimization techniques to jointly explore circuit topologies and optimize device sizing, automatically generating performance-specific, fully sized schematic netlists. AnalogCoder-Pro employs rejection sampling for fine-tuning LLMs on high-quality synthesized circuit data and introduces a multimodal diagnosis and repair workflow based on functional specifications and waveform images. By leveraging LLMs to interpret generated circuit netlists, AnalogCoder-Pro automates the extraction of critical design parameters and the formulation of parameter spaces, establishing an end-to-end workflow for simultaneous topology generation and device sizing optimization. Extensive experiments demonstrate that these orthogonal approaches significantly improve the success rate of analog circuit design and enhance circuit performance.

Problem

• Automating analog circuit design to reduce expert dependency
• Integrating LLMs for joint topology and device optimization
• Enhancing circuit performance via multimodal diagnosis and repair

Innovation

• Multimodal LLM integrates generation and optimization
• Rejection sampling fine-tunes LLMs on circuit data
• Automated parameter extraction and topology generation