德拉姆
收发机
均衡(音频)
校准
相(物质)
光电子学
过程(计算)
剥离(纤维)
材料科学
电气工程
计算机科学
电子工程
物理
工程类
解码方法
电信
CMOS芯片
量子力学
操作系统
作者
Jin‐Hyeok Baek,Janghoo Kim,Yoo‐Chang Sung,Jaewoo Jeong,Jin‐Kwan Park,Hwan Sool Oh,Bo-Hyeon Lee,Dong-Wan Ko,Tae-Won Oh,Sanghyun Hong,Chang-Ki Kwon,Daihyun Lim,Myeong-O Kim,Seung-Jun Bae,Tae-Young Oh,Sang-Jun Hwang
出处
期刊:
日期:2025-02-16
卷期号:: 510-512
被引量:2
标识
DOI:10.1109/isscc49661.2025.10904794
摘要
The rapid growth of AI, AR, and VR applications has increased demand for high-speed data processing and high-density low-power memory modules. LPDDR5X has played an important role in meeting these demands [1]. LPDDR5X introduced optional JEDEC features, such as per-pin decision-feedback equalizers (DFE), offset calibration, pre-emphasis transmitters, and read duty-cycle adjusters (RDCA) [2], to enhance 10 performance; however, LPDDR5X faces various speed-requirement challenges. Recently, there has been a strong demand for high-speed operation beyond 10.7Gb/s/pin; the low-power compression-attached memory module 2 (LPCAMM2) for servers has emerged to address this low-power high-capacity requirement. To achieve sufficient 10 timing margins, using lossy channels, additional transceiver techniques are required. In this paper, we present a 4-phase self-calibration technique to reduce clock phase skew further and three different AC-coupled equalization schemes to increase operating frequency. By utilizing these techniques, we have successfully validated a 16Gb LPDDR5X operating at 1.05V in a 5th-generation 10nm DRAM technology.
科研通智能强力驱动
Strongly Powered by AbleSci AI