NAND flash memory has been widely used for data storage due to its high density, high throughput, and low power. However, as the flash memory scales to smaller process technologies and stores more bits per cell, its reliability is decreasing. The error correction coding can be used to significantly improve the data reliability; nevertheless, the advanced ECCs such as low-density parity-check (LDPC) codes generally demand soft decisions while NAND flash memory channel provides hard-decisions only. Extracting the soft information requires the accurate characterization of flash memory channel and the effective design of coding schemes. To this end, we have presented a novel LDPC-TCM coding scheme for the Multilevel Cell (MLC) flash memories. The a posteriori TCM decoding algorithm is used in the scheme to generate soft information, which is fed to the LDPC decoder for further correction of data bits. It has been demonstrated that the proposed scheme can achieve higher error correction performance than the traditional hard-decisions based flash coding algorithms, and is feasible in the design practice. Further with the LDPC-TCM, we believe it is important to characterize the flash memory channel and investigate a method to calculate the soft decision for each bit, with the available channel outputs. We studied the various noises and interferences occurring in the memory channel and mathematically formulated the probability density function of the overall noise distribution. Based on the results we derived the final distribution for the cell threshold voltages, which can be used to instruct the calculation of soft decisions. The discoveries on the theoretical level have been demonstrated to be consistent with the real channel behaviours. The channel characterization and model provided in this dissertation can enable more design of soft-decisions based ECCs for future NAND flash memories. The data pattern processing algorithm deals with the write patterns and targets to lower the proportion of patterns that would introduce data errors. On the other hand, the voltages applied to the memory cells charges the MOSFET capacitances frequently on programming these data patterns, leading to the power problem. The high energy consumption and current spikes also cause reliability issue to the data stored in the flash memory. This dissertation proposes a write pattern formatting algorithm (WPFA) attempting to solve the two problems together. We have designed and implemented the algorithm and evaluated its performance through both the software simulations and hardware synthesis.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:690981 |
| Date | January 2016 |
| Creators | Xu, Quan |
| Publisher | City University London |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://openaccess.city.ac.uk/15120/ |
Page generated in 0.0015 seconds