⚡ 本页包含 AI 生成的分析内容，仅供参考
该论文提出一种算法拓扑优化框架，能够自动综合非直观的多层宽带毫米波低噪声放大器（LNA），在布局级别同时优化有源和无源器件，从而改善增益、带宽、噪声系数和功耗之间的权衡。通过180nm BiCMOS工艺制造的3级SiGe HBT LNA原型验证了该方法，实现了26.3dB增益、22-46GHz带宽和3.1dB噪声系数。
Abstract An algorithmic topology optimization framework is presented to autonomously synthesize nonintuitive, multilayered wideband mm-Wave LNAs with arbitrary stage count. Cooptimization of actives and passives directly at a layout level enables improved tradeoffs among metrics like gain, bandwidth, NF, and DC power. A 3-stage SiGe HBT LNA prototype fabricated in a 180nm BiCMOS process achieves 26.3dB gain, 22-to-46GHz 3dB bandwidth, 3.1dB NF and 18mW DC power, experimentally validating the method. Inverse design and topology optimization have recently emerged as powerful techniques for autonomously synthesizing mm-Wave amplifiers with nonintuitive pixelated layouts that can significantly outperform classical designs based on lumped or distributed elements such as inductors, transmission lines, capacitors, and transformers. Recent machine learning (ML)-based approaches [1–2] have demonstrated amplifier synthesis using singlelayer pixelated passives, but these result in large chip area
Vinay Chenna, Hossein Hashemi
University of Southern California, Los Angeles, CA