NSF-BSF: CNS Core: Small: Reliable and Zero-Power Timekeepers for Intermittently Powered Computing Devices via Stochastic Magnetic Tunnel Junctions

点击翻译按钮获取中文摘要

This project seeks to advance the state-of-the-art in timekeeping for low-power, battery-free, energy harvesting embedded systems. These systems must perform computation and sensing tasks across multiple, frequent power failures. This intermittent operation requires a strong sense of time to ensure no energy is wasted and computed results are not stale. Existing timekeepers (i.e. real-time clocks) require significant amounts of energy to turn on and operate, are relatively large (millimeter scale), are slow to restart, and use energy proportional to the length of a power failure (which can be impractically long). This project takes advantage of the unique properties of emerging nano-scale non-volatile magnetic random-access memory (MRAM) devices, particularly adjustable data retention length, to create orders of magnitude lower power, adjustable, ultra-small timekeeping devices. A system can set a few bits of the timekeeper and check them later to see which ones have lost information. By tailoring each bit, or array of bits, to particular lengths of time, the system can measure time regardless of whether a power failure occurred. Such a timekeeper requires the same amount of energy if it times an outage of one second, one day, or one year, with arbitrary accuracy depending on the number and characteristics of the magnetic elements used. The project cuts across the system stack to realize the benefits of this timekeeper; exploring circuit designs, architectural, and system paradigms to embed this new timekeeping device in the context of intermittent computing: as an alternative to a runtime clock, an interrupt source, a data and input/output (I/O) controller, and as an expiring memory allocator. These modules are made available to application developers to enhance the timeliness of their programs.The technologies developed in this project will enable more efficient and accurate timekeeping for an important emerging class of computer systems. Batteryless and energy harvesting computing devices offer a more sustainable future for the Internet of Things and enable numerous health, agriculture, and intelligent infrastructure applications. The elimination of batteries in remotely powered sensors will mitigate the growing waste stream resulting from battery wear out and replacement. The fabrication and eventual distribution of the devices and systems from this project will provide a more accurate timekeeping module for future batteryless systems, enhancing the reliability, security, and timeliness of batteryless systems. The resulting tools and experimental results will enable robust cross-stack simulation for battery-free devices and fabrication methods for magnetic memory devices. Outcomes of this project will be integrated into several ongoing outreach activities that expose K-12 and undergraduate students to the usage and application of embedded computing. This effort will train graduate students in essential skills across computing, materials, and physics and engage them in mentoring, teaching, and outreach activities via GoBabyGo!– where students help build power wheelchairs from ride-on toy cars for children with mobility issues.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.