Writing to dirty flash memory

Abstract

Flash memory is a non-volatile computer memory comprised of blocks of cells, wherein each cell can take on $\sl{q}$ different charge levels corresponding to the number of electrons it contains. In the multi-level flash memory, increasing a charge level in a cell is easy, however reducing a charge level forces all the other cells in a block to be erased. This erasing operation is undesirable and need a more cumbersome procedure. And each erasing operation can cause memory endurance problem. Thus we consider designing a novel coding technique to deal with this. The goal is to maximize the number of writing operations before a block erase operation is required. There are many researches about maximizing the number of writing operations in recent years. But they only consider the maximum number of writing operations. They do not consider the writing rate of the multi-level memory for every writing operation. In this thesis, we will consider both the number of rewriting operations and the writing rate in terms of capacity. We first model the channel for multi-level flash memory. We will call this $\it{dirty flash channel}$. And then we will calculate the capacity of our channel on the scope of the information theory. Then by analyzing the capacity region, we will consider the number of writing operations before erasing the block and suggest writing schemes.