A novel method to characterize DRAM process variation by the analyzing stochastic properties of retention time distribution

This study proposes an innovative method to measure the variation of cell leakage current. Extreme cell leakage determines DRAM refresh time (tREF). Although the average leakage current from the test element group (TEG) has been the only index for determining cell leakage, it does not provide the distribution of unit cell leakage. We find that cell leakage distribution can be calculated from the slope at the retention time-fail bit plot. A steep slope indicates a small cell leakage distribution that corresponds to a long tREF. It is proved with statistical models and experimental results with mass data.