COmPOSER: Open-source tool for Circuit Optimization of mm-wave/RF circuits

What is it about?

Radio-frequency (RF) and millimeter-wave (mm-wave) circuits are essential to modern wireless communication systems, including 5G/6G networks, satellite communication, and high-speed wireless devices. Designing these circuits is particularly challenging because their performance depends not only on transistor and circuit parameters but also on the physical arrangement of components and interconnections on a chip. At high frequencies, electromagnetic (EM) effects can significantly alter circuit behavior, making conventional schematic-level optimization insufficient. Engineers therefore spend considerable time iterating between circuit design, physical layout, and EM simulations to achieve the desired performance. This paper introduces COmPOSER, an open-source, end-to-end RF/mm-wave design automation framework that translates circuit performance specifications into optimized, physically realizable chip layouts. Unlike approaches that optimize circuits independently of their physical implementation, COmPOSER integrates circuit optimization and layout generation within a unified workflow. It combines physics-based analytical equations for active devices with machine-learning-based EM models for passive components, such as inductors and transmission lines, enabling fast and physically informed circuit optimization. The framework automates transistor and passive-component sizing, layout generation, placement, routing, power-delivery network synthesis, and post-layout verification. Evaluated on low-noise amplifiers (LNAs) and power amplifiers (PAs) operating up to 60 GHz in a commercial 65 nm CMOS technology, COmPOSER achieves performance comparable to expert-designed circuits while improving design productivity by approximately 100–300×. Importantly, COmPOSER is publicly available as an open-source framework, enabling researchers and engineers to examine, reproduce, and extend the proposed automation methodology.

Featured Image

Photo by Immo Wegmann on Unsplash

Photo by Immo Wegmann on Unsplash

Why is it important?

As wireless technologies advance toward higher frequencies, RF/mm-wave circuit design is becoming increasingly complex and time-consuming. A fundamental challenge is the strong interaction between circuit parameters and physical layout. At mm-wave frequencies, even small interconnect parasitics and electromagnetic coupling can substantially change circuit performance. Our experiments demonstrate that ignoring these effects can shift a circuit designed for 60 GHz operation to 40–45 GHz, accompanied by significant degradation in gain, noise, and output power. COmPOSER addresses this problem by bringing EM-aware circuit optimization and automated physical design into a unified framework. Its key innovation is combining fast, physics-based circuit equations with accurate machine-learning models of EM-sensitive components, enabling efficient optimization without requiring expensive EM simulations at every step. The framework further automates physical layout generation, including RF-aware placement, routing, and power-delivery network synthesis, followed by independent EM and circuit-level verification. The results demonstrate that this approach is practical: COmPOSER successfully synthesizes seven LNA and PA designs with post-layout performance comparable to expert designs. For a representative LNA, the design process is reduced from approximately 70 hours of manual effort to about 24 minutes, including verification. A particularly important contribution is that COmPOSER is open-source. RF/mm-wave design automation research often depends on specialized tools, technology-specific models, and complex implementation workflows, making it difficult for others to reproduce or build upon published methods. By publicly releasing COmPOSER, we aim to lower these barriers and provide researchers with a practical foundation for evaluating new optimization algorithms, incorporating additional circuit architectures, and advancing physical-design-aware automation. This makes the contribution valuable not only as a demonstrated design methodology but also as a reusable research platform for the broader electronic design automation community.

Perspectives