| 研究実績の概要 |
We have finished performing variability and reliability analysis of write stability in silicon-on-thin-BOX (SOTB) SRAM cells at low VDD.
Firstly, in order for a good yield estimation, four write stability methods (write butterfly curve, write N-curve, bit-line method, and word-line method) are compared. It is found that the latter two methods are good candidates. In addition, the reason why write noise margin (WNM) of write butterfly curve deviatse from a normal distribution is clarified.
Besides, the impact of random telegraph noise (RTN) is discussed. Based on our experimental results, a statistical model is proposed and it is concluded that RTN degrades Vmin - lowest VDD that SRAM can operate - by 48 mV at capacity of around 6.3 sigma, which should be paid attention to in SRAM design.
|
| 今後の研究の推進方策 |
The following work will go to details about write stability metrics in silicon-on-thin-BOX (SOTB) SRAM cells at low VDD. There are two main problems as follows.
Firstly, the reason why write noise margin (WNM) from write N-curve deviates from a normal distribution is still not clear. Assuming that two modes exist, cell transistors' conditions are to be compared for each mode and the distribution pattern for each mode is to be plotted.
Secondly, a new write stability metric based on our present work is to be proposed. Write butterfly curve is a popular method among engineers due to its simplicity though with its own drawbacks. Based on our understanding, a new method is to be proposed to replace conventional write butterfly curve for yield estimation at low VDD.
|
