I-Corps Teams: Machine Learning (ML)-powered Data Analyzer for Radio Frequency Integrated Circuits (RFIC) Design

I-Corps Teams：用于射频集成电路 (RFIC) 设计的机器学习 (ML) 驱动的数据分析器

• 批准号: 2016398
• 负责人: Roman Lubynsky
• 金额: $ 5万
• 依托单位: Massachusetts Institute of Technology
• 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: 2020-02-01 至 2021-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2016398&HistoricalAwards=false
• 关键词: Corps; Teams; Machine; Learning; ML

The broader impact/commercial potential of this I-Corps project is the development of a software platform for radio frequency integrated circuits (RFIC) design with more accurate simulation prediction. The RFIC industry, key for wireless systems, is approximately $50 B, and RFIC companies spend about $15 B on the final trial-and-error fabrication process known as "tape-out". Each tape-out takes 12 weeks and costs ~$1 M, and 3-5 may be required. Significant savings may be realized in both time and cost by reducing design iterations.This I-Corps project is based on the development of RFICs . RFIC’s parasitic effect and poor simulation accuracy requires multiple tape-outs to meet specifications. In the future 5G era, this problem gets worse as mmWave frequency parasitics are even harder to model and more tape-out rounds will be required. Using the technology under development, it may be possible to improve significantly simulation accuracy by combining traditional physics-based models and customized machine-learning (ML) training in RFIC simulation.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.

该 I-Corps 项目更广泛的影响/商业潜力是开发用于射频集成电路 (RFIC) 设计的软件平台，该平台具有更准确的仿真预测。RFIC 行业是无线系统的关键，其成本约为 50 B 美元，而 RFIC 的成本约为 50 美元。公司在最终的试错制造过程（称为“流片”）上花费约 15 B 美元，每次流片需要 12 周，成本约为 100 万美元，可能需要节省 3-5 美元。通过减少设计迭代，可以在时间和成本上实现。这个I-Corps项目是基于RFIC的寄生效应和较差的仿真精度而需要多次流片才能满足未来5G时代的规范。由于毫米波频率寄生更加难以建模，并且需要更多的流片轮次，因此情况会变得更糟。使用正在开发的技术，通过结合传统的基于物理的模型和定制的机器学习，可以显着提高模拟精度。 (ML) RFIC 模拟培训。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。