SBIR Phase I: High performance non-volatile-memory circuits for artificial intelligence

SBIR第一阶段：用于人工智能的高性能非易失性存储器电路

• 批准号: 2014959
• 负责人: Yu Lu
• 金额: $ 22.5万
• 依托单位: SUPERMEM INC.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-05-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2014959&HistoricalAwards=false
• 关键词: SBIR; Phase; performance; non; volatile

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to enable wider adoption of Artificial Intelligence (AI) through development of a high-bandwidth, low-cost integrated circuit memory solution. AI is poised to make fundamental changes to how people live and work but require dramatic improvement in computing power and efficiency. So-called "edge AI" applications, such as smart home devices, smart city, and autonomous vehicles, will demand on-device AI subject to tight power and cost constraints. Current configurations with conventional memories may not offer a combination of performance, cost, power, operating temperature range necessary for many edge applications. The proposed project will advance the development of a new circuit architecture to address these challenges. This Small Business Innovation Research Phase I project develops high-speed analog circuit architectures and techniques in the context of emerging memory readout circuits. The proposed novel read circuit is applicable to magnetoresistive random-access memory (MRAM) and other emerging memories to achieve high performance and low power. This Phase I project will fully develop the concept and overcome key technical challenges that include (1) increasing differential amplifier speed and margin under severe area limitation, (2) reducing noise level and silicon area of a novel signal passing circuit, and (3) searching for an efficient way to serve typical data access requests with the novel read operation. The project goals are to validate the novel MRAM read circuit with silicon test data and perform a model study to quantify system level benefits.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.

这项小企业创新研究（SBIR）第一阶段项目的更广泛影响/商业潜力是通过开发高带宽、低成本集成电路存储器解决方案，使人工智能（AI）得到更广泛的应用。人工智能有望从根本上改变人们的生活和工作方式，但需要大幅提高计算能力和效率。所谓的“边缘人工智能”应用，如智能家居设备、智能城市和自动驾驶汽车，将需要受到严格的功率和成本限制的设备上人工智能。传统存储器的当前配置可能无法提供许多边缘应用所需的性能、成本、功耗和工作温度范围的组合。拟议的项目将推动新的电路架构的发展，以应对这些挑战。这个小型企业创新研究第一阶段项目在新兴存储器读出电路的背景下开发高速模拟电路架构和技术。提出的新型读电路适用于磁阻随机存取存储器（MRAM）和其他新兴存储器，以实现高性能和低功耗。这个第一阶段的项目将充分发展这个概念，并克服关键的技术挑战，包括(1)在严重的面积限制下提高差分放大器的速度和裕度，(2)降低噪声水平和新型信号传递电路的硅面积，以及(3)寻找一种有效的方法来服务典型的数据访问请求与新型读取操作。该项目的目标是用硅测试数据验证新型MRAM读电路，并进行模型研究以量化系统级效益。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。