⚡ 本页包含 AI 生成的分析内容,仅供参考
本文提出了一款基于2nm工艺的全数字LPDDR6 PHY,通过四相时钟架构与多级FIFO推测式DFE,实现了14.4Gb/s/pin的高速数据传输,同时降低了抖动敏感度和时序约束。采用时钟门控的高效模式显著降低了功耗。
LPDDR6 PHY that achieves 14.4Gb/s/pin, increasing bandwidth while optimizing power. Key innovations include quadrature clocking, with QEC, to reduce jitter sensitivity, and a multi-level FIFO-based speculative DFE to relax timing constraints in standard-cell implementation. An efficiency mode, employing clock gating, dramatically reduces power. A reconfigurable serializer supports both 1:6 and 1:8 DFIs, ensuring high bus efficiency and backward compatibility. Bandwidth demands of low-power double-data-rate (LPDDR) interfaces continues to grow to support AI, augmented reality (AR), and high-resolution multimedia workloads [1-3]. To meet demand, the PHY’s internal clock frequency is aggressively scaled; resulting, in
Yoonjae Choi, Daero Kim, Myunggon Kim, Sangmin Lee, Hansol Kang, Gwangwon Lee, Yoonhyung Kim, Sungsik Kang, Hyun Cho, Boyoung Kang,
Kyoungwon Lee, Jaegeun Song, Jinho Choi, Shinyoung Yi, Billy Koo, Kwanyeob Chae, Hyo-Gyuem Rhew Samsung Electronics, Yongin, Korea Abstract This paper presents a 2nm all-digital