LLMs on the Path Toward Analog IC Automation: ChatGPT as an ADC Design Expert
Year: 2026
Type: conference-paper
Abstract: Numerous machine learning techniques have been applied to address complex problems in analog integrated circuit (IC) design. However, the highly non-linear nature of analog sub-blocks makes system-level automation particularly challenging due to the lack of robust circuit models. Such automation can be improved by integrating nonlinear design parameters with the insights of experienced designers who have developed these skills over decades. In this context, large language models (LLMs) have the potential to assist designers in complex analog design tasks, particularly during early-stage architecture selection. In this paper, we present an evaluation framework to assess ChatGPT’s capability to provide design guidance for analog circuit problems, particularly analog-to-digital converters (ADCs). The model is prompted using a dataset-anchored approach, where Boris Murmann’s ADC survey serves as the reference dataset. Its performance is evaluated through twenty-four design scenarios across four difficulty levels. The results highlight the potential usage of LLMs as effective assistants for early-stage analog IC design exploration through design-tailored prompting.
Institutions: Gebze Technical University, Boğaziçi University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203624924