CAREER: Ultra-Low Power Programmable Analog Signal Processing Systems

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Demand for low-power, sophisticated information processing and control systems grows daily. Consumers seek hand-held devices delivering more computationally intensive capabilities in a smaller space with a longer battery life. Industry requires reduced operating costs by reducing the power consumed by installed systems. Military applications demand deployable smart sensors with greater mission longevity, and that sustain themselves on minimal resources. Such demands have driven recent research activity to reconsider analog embedded processing. Advances in analog floating-gate circuits enable complex algorithms to be realized in the analog domain in a manner that is both programmable and adaptive. With scalability similar to CMOS circuits, massively parallel analog processing becomes feasible. The focus of this CAREER research in cooperative analog-digital signal processing (CADSP) is on freedom of movement of the partition between the analog and digital systems. By allowing more of the computation to be performed in low-power analog circuits, cooperative analog-digital signal processing enables advanced functionality in very low-power embedded systems and smart sensors. This project pioneers the design and implementation of embedded signal processing and control systems that perform significant processing in analog as well as digital circuits. The analog circuits are extremely small and low power, making it possible to investigate entirely new methods of computing, exploiting large-scale parallelism and redundancy in a very small space. This entails a very high-density field-programmable analog array (FPAA) based on floating-gate analog circuits. The FPAA device operates in tandem with a traditional digital field-programmable gate array (FPGA). Significant impact is expected for signal processing and control applications, enabling low-power analog FPAA processing for computations previously viewed as feasible only in the digital domain. The integration of research and education provides new theoretical analyses, software tools, silicon chips, and education materials to aid both academics and industrial designers. Broader impact is sought through tools that are made freely available and on-line education materials, disseminated via university courses and short courses.