An algorithm that can design RF chips in days instead of months. The resulting layouts look like QR codes and outperform conventional approaches.
Researchers at the USC Viterbi School of Engineering have developed an algorithmic framework for designing radio frequency integrated circuits (RFICs), enabling the discovery of unconventional chip layouts that traditional design approaches may not reach. Developed by PhD student Vinay Chenna under the guidance of Hossein Hashemi, the framework is aimed at accelerating the development of wireless technologies used in devices such as smartphones, automotive radar systems, satellites, and Wi Fi equipment.
The approach focuses on automating RF chip design, a process that traditionally relies heavily on engineering intuition and iterative manual refinement. Instead of starting from predefined circuit architectures, the framework divides the chip into a three dimensional grid across multiple metal layers and continuously modifies the layout through simulation driven optimization.
One of the key benefits of the framework is its ability to reduce chip design timelines from several months to a matter of days or weeks. The method also explores significantly larger design spaces than conventional approaches, allowing it to identify layouts that may not be intuitive to human designers. According to the researchers, the resulting geometries often resemble QR code like patterns rather than conventional chip structures.
Unlike many AI assisted chip design approaches that depend on machine learning and large training datasets, the framework uses metaheuristic algorithms based on guided randomness and iterative physics simulations. This allows the system to operate without pre existing design databases while handling highly complex multilayer chip architectures.
The generated designs have already been fabricated using standard semiconductor manufacturing processes through Tower Semiconductor. The researchers demonstrated the approach on components including power amplifiers, low noise amplifiers, and signal filters for high frequency wireless systems.
Hossein Hashemi in the Ming Hsieh Department of Electrical and Computer Engineering of Viterbi, says, “At this stage, some of these geometries are difficult for us to fully explain, but they are demonstrating performance levels that conventional approaches have struggled to achieve.”
