Circuit- and Signal-Domain Co-Design in GaN Technologies for 6G Wireless Systems

About This Grant

Cellular wireless communications are an integral part of modern life, with the demand for high data-rate communications continuously increasing. To support this need, the cellular infrastructure in the United States uses a combination of legacy frequency bands from 4G standards as well as newly allocated spectrum for 5G/6G systems. As a result, a cellular base station must support a variety of frequency ranges and standards, often simultaneously in the case of carrier aggregation. Conventional industry solutions use a single radio-frequency (RF) hardware chain to serve each band, for example using separate dedicated hardware for the two Advanced Wireless Services 1 (AWS-1) bands of 1805-1880 MHz and 2110-2155 MHz even though these frequencies are relatively close to each other. The RF front-end hardware therefore exists in duplicate or triplicate, depending on the range of frequencies to be covered. If a single transmitter could operate instead over the entire frequency range of 1805-2155 MHz, this would cut in half the size, cost, complexity, and capital investment for the cellular base stations. At the same time, a successful broadband solution must not incur any energy efficiency penalty, as a reduction of even five percentage points is prohibitive in terms of operating costs and the consequent need for heat removal. This efficiency requirement precludes the use of existing hardware solutions for bandwidth extension, which are well-known to exhibit a bandwidth-vs-efficiency tradeoff. This project will develop alternative solutions involving both new hardware design and machine learning (ML) algorithms to address these challenges. The project will also train an engineering workforce able to solve challenges from a multidisciplinary perspective. The aim of this project is to address the growing demand for high-power RF transceiver front-ends capable of operating over multiple bands without incurring any energy efficiency penalty through the co-design of RF transceiver hardware and signal processing software. The majority of previous efforts addressing the bandwidth-vs-efficiency tradeoff have approached this problem from a purely circuit design perspective. Similarly, linearization of, and signal generation for, cellular RF transceiver front-ends falls squarely in the domain of signal processing design. The research of this project, on the other hand, approaches the challenge from the basis that a true wideband and efficient solution cannot arise from a narrow technical approach, but instead requires a cross-disciplinary solution incorporating both hardware design and artificial intelligence (AI)/ML-assisted signal processing that can control the data signals to be transmitted in a way that optimizes both spectral efficiency and energy efficiency. The new research approach of this project eliminates the artificial boundary between hardware and software in conventional designs which typically pre-determine the signal generation strategy for a dual-drive power amplifier. By allowing the signal processing and separation to be fully two-dimensional, an additional degree of design freedom is enabled. The project exploits this new 2D design space for both hardware design and signal processing to optimize bandwidth and efficiency. Because of the complexity of the function space, optimization tools are employed. The end goal is a co-design of hardware and signal processing. Beyond these technical goals, the project supports US workforce development for the wireless industry by training RF engineers with critical interdisciplinary skills in both hardware design and AI/ML tools for the next-generation wireless communication systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.