SHF: Small: Printed Computer Systems

虽然计算在当今的社会和经济中似乎无处不在，但大量重要的领域仍然被最少涉及。例如，考虑一下价值超过10万亿美元的快速消费品(FMCG)市场。诸如包装食品、饮料、洗漱用品、非处方药和其他消耗品等一次性产品的销售基本上没有任何嵌入式计算设备，这些设备可以帮助识别和跟踪、质量监控、品牌认证或互动。快速消费品领域以及低端医疗保健(例如绷带和伤口敷料)、农业和环境等领域没有看到太多计算渗透的主要原因是今天基于硅的计算系统的成本限制。硅基系统的成本仍然远远高于这些领域的成本要求。例如，几种快速消费品产品的物品级标签必须具有与条形码相同的成本，而基于硅的系统由于此类系统的制造、测试和组装成本较高，无法满足这一要求。印刷电子学已经成为瞄准这些应用领域的有前途的技术。印刷技术通常依赖于无掩膜、便携和添加的制造方法，这些方法可以大大降低成本和生产时间表。这样的技术还带来了顺应性和无毒的设备。这项工作将集中在使这种方法廉价、可靠和灵活。结果可以确保对制造业经济的所有组件和部分进行电子标记。学生将在建筑、电路和材料科学的交叉点上接受独特的、新兴的跨学科领域的培训。以前关于印刷有源电子的工作主要集中在制造简单的电路上。以前的工作没有探索印刷计算机系统的体系结构、设计或优化空间，例如微处理器和加速器。这并不令人惊讶，因为这样的探索需要最近才开始提供的过程设计工具包(PDK)。综合和物理设计标准单元库是进行建筑探索的先决条件。该项目将为电解液门控FET和碳纳米管薄膜晶体管印刷技术开发标准单元库，并通过制造测量进行校准。这些库(将公开以供更广泛的使用)对于打印计算机系统的设计空间探索至关重要。研究人员将开发和制作印刷计算机系统的不同微体系结构构件的原型。待开发和原型的构建块包括用于印刷微处理器的指令和数据存储器、ALU等，以及用于印刷机器学习加速器的决策树和感知器。然后，这些构件将与开发的标准单元库结合使用，以探索印刷计算机系统的设计空间。探索的预期结果包括印刷微处理器和加速器的指令集体系结构和微体系结构设计原则，以及将印刷硬件的面积和功耗降至最低的优化。总体而言，这项研究将有助于使硬件制造民主化，使计算即使在一次性超低成本领域也变得无处不在。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

While computing appears to be ubiquitous in today's society and economy, a large number of important domains are still minimally touched. Consider the over 10-trillion dollar fast-moving consumer goods (FMCG) market, for example. Disposables such as packaged foods, beverages, toiletries, over-the-counter drugs, and other consumables are sold largely without any embedded computing devices that could help with identification and tracking, quality monitoring, brand authentication, or interactivity. The primary reason why FMCG domains as well as domains such as low-end healthcare (e.g., bandages and wound dressings), agriculture, and environment have not seen much penetration of computing is the cost limitation of today's silicon-based computing systems. Silicon-based systems continue to cost much more than the cost requirements of these domains. For example, item-level tagging of several FMCG products must have costs equivalent to that of a barcode, which silicon-based systems cannot meet due to the high manufacturing, testing, and assembly costs of such systems. Printed electronics has emerged as promising technology to target such application domains. Printing technologies often rely on maskless, portable, and additive manufacturing methods which can greatly reduce costs and production timelines. Such technologies also lead to devices that are conformable and nontoxic. This work will focus on making this method cheap, reliable and flexible. The outcomes can ensure electronic tagging of all components and parts of the manufacturing economy. Students will be trained in a unique and up-and-coming interdisciplinary area at the intersection of architecture, circuits, and materials science. Prior work on printing active electronics has largely focused on fabricating simple circuits. No prior work has explored the architecture, design, or optimization space of printed computer systems, e.g., microprocessors and accelerators. It is not surprising since such exploration requires Process Design Kits (PDKs) that have started becoming available only recently. A synthesis and physical design standard cell library is a pre-requisite for an architectural exploration. This project will develop standard cell libraries for Electrolyte Gated FET and CNT-thin film transistor printed technologies calibrated with fabrication measurements. These libraries (to be made public for broader use) are critical for design space exploration of printed computer systems. The investigator will develop and prototype different microarchitectural building blocks of a printed computer system. Building blocks to be developed and prototyped include instruction and data memory, ALUs, etc., for printed microprocessors, and decision trees and perceptrons for printed machine learning accelerators. These building blocks will then be used, in conjunctions with the developed standard cell libraries, to explore the design space of printed computer systems. Expected outcome of the exploration includes instruction set architecture and microarchitectural design principles for printed microprocessors and accelerators, and optimizations to minimize area and power of printed hardware. Overall, the research will help democratize hardware fabrication making computing ubiquitous even in disposable ultra-low-cost domains.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.