NAND flash sees continued growth in mass storage
Industrial storage applications are a significant contributor to the shift of storage based on NAND flash technology. In the industrial market, compact size, low power and the inherent robustness of a technology with no moving parts are major benefits of flash storage. Furthermore, the continuing growth of the Industrial Internet of Things (IIoT) powering smart factories, and other industrial solutions like 5G networks and smart grids have resulted in an increasing demand for NAND flash storage.
But why NAND flash? What are its advantages over other types of storage.
What is NAND flash?
What differentiates flash from other types of memory is the way the data is stored. In DRAM and SRAM, the data is lost when power is removed.
In flash memory, data is stored in a modified field effect transistor (FET). A normal FET has a metal “gate” electrode that sits above a conducting channel, separated from it by an insulating layer. The gate voltage controls how much current can flow through the transistor. In a flash memory cell, there is an extra electrically-isolated gate buried in the insulting layer. This “floating” gate can have charge injected or removed by forcing electrons to tunnel through the insulation from the external gate. The extra charge means that the transistor will turn on when the memory is read, resulting in a 0 output.
Because the floating gate is electrically isolated, the charge will remain there, even if power is removed. This gives flash memory its main advantage: non-volatility.
The cells in a column of the array are connected in series, in a similar way to a NAND gate.
Each row of cells is called a “page” and the whole array forms a “block”. The entire flash memory will be made up of several blocks.
An important characteristic of NAND flash is that data is written a page at a time. Writing can add charge to cells but cannot remove it, meaning that only ‘0’s can be written. So, to enable arbitrary bit values to be written, data must always be written to an erased page.
However, cells can only be erased at the block level. As a result, the controller that interfaces the host to the flash storage needs to use some very sophisticated algorithms to efficiently manage the use of pages and blocks.
Multi-level cells and 3D flash
In the description above, we have considered the case where each cell stores just a 0 or 1. To increase storage density, flash memory designers moved beyond this single level cell (SLC) by using four different levels of charge. This multi-level cell (MLC) is equivalent to storing two data bits in each memory cell, thus doubling the storage density. This has been extended to store three bits in a triple-level cell (TLC) and four bits in a quad-level cell (QLC). Note that the names are slightly misleading, as they refer to the number of bits, not the number of actual charge levels. A QLC requires 16 different levels of charge to be programmed and measured.
With growing demand for storage density and bandwidth, flash memory makers have now decided that the only way is up. The latest flash memories are based on 3D technology. Rather than just a 2D array of memory cells on the surface of a silicon chip, multiple layers (64 or more) are used to create a thee-dimensional structure of memory cells in the silicon. Despite the manufacturing complexity of 3D memory, the advantages in storage density, and hence reduced cost per bit, make it worthwhile.
Conclusion
NAND flash is still an evolving technology that has the ability to meet future demands for mass storage in terms of power efficiency, data density and cost per bit. The lower storage costs enabled by packing more bits per cell, and more cells per chip using 3D technology, is making SSDs increasingly competitive with hard drives. Combined with lower power consumption and increased reliability when paired with the right flash memory controller, NAND flash will increasingly be the storage solution of choice in an ever widening range of applications.