Abstract: Methods and systems for improving the reliability of data stored within a semiconductor memory are described. One issue with determining stored data states for memory cells within a NAND-type memory is that the voltage at the source end of a NAND string may vary greatly from when a memory cell of the NAND string is program verified to when the memory cell is subsequently read leading to bit errors. To compensate for this variability in the source line voltage, different sensing conditions (e.g., the bit line voltages and/or the sensing times) may be applied during a read operation to different sets of memory cells depending on the source line resistance from the memory cells or on the source line voltage zone assigned to the memory cells.

Abstract: Methods and systems for improving the reliability of data stored within a semiconductor memory are described. One issue with determining stored data states for memory cells within a NAND-type memory is that the voltage at the source end of a NAND string may vary greatly from when a memory cell of the NAND string is program verified to when the memory cell is subsequently read leading to bit errors. To compensate for this variability in the source line voltage, different sensing conditions (e.g., the bit line voltages and/or the sensing times) may be applied during a read operation to different sets of memory cells depending on the source line resistance from the memory cells or on the source line voltage zone assigned to the memory cells.

Abstract: Techniques are provided for optimizing the programming of memory cells by obtaining a metric which indicates a program or erase rate of the memory cells. In one approach, a count of pulses used to program the cells to different verify levels of respective data states is stored. A slope of a straight line fit of data points is then obtained. Each data point can include one of the verify levels and a corresponding one of the counts. An optimal step size is determined based on the slope. The counts may exclude one or more initial program voltages while the cells are programmed sufficiently to allow faster and slower cells to be distinguished, e.g., in a natural threshold voltage distribution. An erase depth can also be adjusted. The cells can be programmed in a separate evaluation or during programming of user data.

Abstract: Techniques are provided for optimizing the programming of memory cells by obtaining a metric which indicates a program or erase rate of the memory cells. In one approach, a count of pulses used to program the cells to different verify levels of respective data states is stored. A slope of a straight line fit of data points is then obtained. Each data point can include one of the verify levels and a corresponding one of the counts. An optimal step size is determined based on the slope. The counts may exclude one or more initial program voltages while the cells are programmed sufficiently to allow faster and slower cells to be distinguished, e.g., in a natural threshold voltage distribution. An erase depth can also be adjusted. The cells can be programmed in a separate evaluation or during programming of user data.