SHF: Small: Bayesian Model Fusion: A Statistical Framework for Efficient Validation and Tuning of Complex Analog and Mixed Signal Circuits

自适应自愈是一种新兴的方法，用于对抗纳米级工艺变化的有害影响，以保持模拟和混合信号(AMS)电路的积极扩展。如今，每个可调的AMS电路都变成了一个大规模、复杂的系统，可以随着时间的推移而自适应地变化。随着器件的持续缩小和关键工艺波动的相对幅度持续增长，与这种复杂系统的硅前验证和硅后调整相关的高得令人望而却步的成本是一个日益严重的问题。因此，迫切需要开发新的统计方法，以最大限度地减少针对未来技术世代的纳米级AMS电路的验证和调谐成本。该项目开发了一种新的统计框架，称为贝叶斯模型融合(BMF)，旨在将自愈AMS电路的硅前验证和硅后调整的模拟和/或测量成本降至最低。建议的BMF技术的动机是，当今的AMS设计周期通常跨越多个阶段(例如，原理图设计、布局设计、第一次流片、第二次流片等)。其关键思想是重用在早期阶段收集的模拟和/或测量数据，以便于在后期以最少的数据量对AMS电路进行有效的验证和调整。它为面向未来IC技术的下一代AMS设计提供了基础设施。拟议的项目提供了一种基于贝叶斯推理的全新AMS设计方法。预计它将在从消费电子到医疗器械的广泛应用中为先进电路带来显著的性能改进。因此，拟议的BMF框架的成功发展将对半导体业产生短期和长期影响。此外，与该项目相结合的教育活动为大学生和工业工程师提供了许多独特的培训机会。它将极大地改善教育基础设施，培养该领域的高素质研究人员和从业人员。

Adaptive self-healing is an emerging methodology to combat the deleterious effects of nanoscale process variations, to maintain the aggressive scaling of analog and mixed-signal (AMS) circuits. Nowadays, each tunable AMS circuit becomes a large-scale, complex system that can adaptively vary over time. The prohibitively high cost associated with pre-silicon validation and post-silicon tuning of such a complex system is a growing problem as devices continue to shrink and the relative magnitude of critical process fluctuations continues to grow. Hence, there is an immediate need to develop new statistical methodologies that minimize the validation and tuning cost of nanoscale AMS circuits for future technology generations. This project develops a novel statistical framework, referred to as Bayesian Model Fusion (BMF), that aims to minimize the simulation and/or measurement cost for both pre-silicon validation and post-silicon tuning of self-healing AMS circuits. The proposed BMF technique is motivated by the fact that today's AMS design cycle typically spans multiple stages (e.g., schematic design, layout design, first tape-out, second tape-out, etc). The key idea is to reuse the simulation and/or measurement data collected at an early stage to facilitate efficient validation and tuning of AMS circuits with a minimal amount of data required at the late stage. It provides a fundamental infrastructure that enables next-generation AMS design for future IC technology.The proposed project offers a radically new AMS design methodology based on Bayesian inference. It is expected to yield significant performance improvement for advanced electrical circuits in a broad range of applications, from consumer electronics to medical instruments. Hence, successful development of the proposed BMF framework will have both short-term and long-term impacts on the semiconductor industry. In addition, the education activities integrated with this project offer a number of unique training opportunities to both university students and industrial engineers. It will substantially improve the education infrastructure and generate high-quality researchers and practitioners in the field.