Generation of accurate reference current for data sensing in high-density memories by averaging multiple Pairs of dummy cells

Methods to generate an accurate reference current by averaging multi-pair dummy cells' currents for distinguishing the data in sense amplifiers (S/As) of a large scale memory with resistance change cell is presented and analyzed. The predicted characteristics are confirmed by comparing them with measurement results of the functionalities and the retention time distributions in a floating body random access memory (FBRAM). The methods are found to be especially effective in situations where signals are seriously degraded such as in sensing the signals of tail bit cells in retention time distributions, making the retention time performance of the FBRAM improved drastically. The sense amplifiers which can accommodate the dummy cell averaging methods are identified to find a necessary condition for a S/A to afford the methods.