Worst-Case Performance Analysis Under Random Telegraph Noise Induced Threshold Voltage Variability
Worst-Case Performance Analysis Under Random Telegraph Noise Induced Threshold Voltage Variability
复制标题
随机电报噪声引起的阈值电压变化下的最坏情况性能分析
DOI：
10.1109/patmos.2018.8464147
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
H. Onodera
中科院分区：
文献类型：
--
作者：
A. Islam;H. Onodera
RTN引起的阈值电压分布有一个长尾，可以严重恶化的最坏情况下的分布。在本文中，我们分析了RTN对最坏情况下的性能的基础上，从65纳米薄体硅低阈值电压过程中提取的变异性模型的影响。基于蒙特卡罗的仿真结果表明，随着电源电压的降低，RTN可以降低最坏情况下的延迟超过10%的关键路径的数量为10。如果关键路径数增加到100，则最坏情况下的延迟退化可高达100%。由于RTN引起的阈值电压波动，在近阈值/亚阈值操作时出现一些异常值。考虑到RTN幅度在弱反型操作时会增加，低压操作需要仔细考虑RTN。
RTN induced threshold voltage distribution has a long tail that can degrade the worst-case distribution severely. In this paper, we analyze the effect of RTN on worst-case performance based on variability models extracted from a 65 nm silicon-on-thin-body low threshold voltage process. Monte Carlo based simulation results reveal that with the lowering of supply voltage, RTN can degrade the worst-case delay by more than 10 % when the number of critical paths is 10. The worst-case delay degradation can go as high as 100 % if the critical path number increases to 100. Because of the RTN induced threshold voltage fluctuation, several outliers appear at near/sub-threshold operation. Considering RTN amplitude can increase at weak-inversion operation, low-voltage operation needs careful consideration of RTN.