A brief history and future perspectives on sizing and layout synthesis of analog/RF integrated circuits

Analog/mixed-signal/RF integrated circuits (ICs) are unavoidable in all applications where electronics interface with the physical world. But whereas digital circuits today are largely synthesized through EDA software, surprisingly, the design of analog circuits in industry still uses a lot of handcrafting, resulting in a low productivity, long and error-prone design cycles, and high development costs. Yet, analog design automation has been an active research area over the past decades. The first part of this paper gives a bird’s eye overview of major research developments in the field of sizing and layout synthesis of analog, mixed-signal and RF integrated circuits. The huge progress in model-based optimization techniques and especially the rebirth of AI and machine learning methods, such as reinforcement learning, have boosted the field. Still, there are shortcomings that limit widespread usage in industry. The second part of this paper presents perspectives on future developments expected in the field. Getting good results with the tools depends on entering proper constraints and design guidance; hence finding better ways to extract and employ these, as well as to explain the results generated by the tools will be key. The recent rise of generative AI and LLM-based methods also creates a whole new spectrum of techniques to assist and automate analog design, promising to obtain better results faster. Finally, topological circuit synthesis and more formal analog/mixed-signal design verification are areas that need to be addressed more fundamentally. The ultimate question is how analog IC design will evolve and how human designers and EDA tools will co-operate in the future.