Most NAND flash SSDs use floating gate cells to store data, but some manufacturers are turning to charge trap cells in an attempt to achieve better endurance and scalability. Drives based on charge trap technologies are less susceptible to damage and leakage. They also consume less energy and are faster to program.
However, charge trap cells have their own challenges, particularly when it comes to reliability. It's anyone's guess where the industry will land.
The floating gate dilemma
Flash drives have been using floating gate cells since their inception. Each cell contains one floating gate that's integrated into the cell's structure. The floating gate traps electrons when voltage is applied to the cell in a specific way. It also releases electrons when voltage is applied in a different way.
In a single-level cell drive, when a floating gate contains electrons, it's considered charged, or programmed, and the cell's bit value is registered as zero. Otherwise, the cell is considered uncharged, or erased, and its bit value is registered as one. The calculations are more complex for multi-level cell and triple-level cell drives, but the fundamentals are the same.
This article is part of
Inside the cell, an oxide layer separates the floating gate from the silicon substrate where the voltage travels into and out of the cell. The oxide layer is thin enough for electrons to pass between the floating gate and substrate when voltage is applied. During a program, or write, operation, electrons move into the floating gate. During an erase operation, electrons move out of the floating gate.
Each program/erase cycle slightly damages the oxide layer, and after a sufficient number of P/E cycles, the oxide layer will erode enough for electrons to start leaking out of the floating gate until it can no longer hold a charge and the cell becomes unusable.
As cell sizes shrink and more bits are packed into each cell, the drives become even more susceptible to damage. Technologies such as wear leveling can prolong the drive's life, but the cells will eventually fail.
Storage analyst Eric Slack talks about how NAND flash degrades over time and becomes less reliable.
Charge trap to the rescue
To address the limitations of floating gate cells, some flash manufacturers, such as Samsung, SanDisk, SK Hynix and Toshiba, are building drives that use charge trap cells. Charge trap cells have been around for a while, but it wasn't until 3D flash came along that vendors started looking seriously at this technology for enterprise-grade SSDs.
In many respects, charge trap cells work much like floating gate cells, with different voltage patterns moving electrons into and out of a trapping layer. But there's one difference. The floating gate uses polycrystalline silicon to provide a conductor for trapping the electrons. The charge trap uses silicon nitride to provide an insulator.
Silicon nitride is less susceptible to defects and leakage than the floating gate, and it requires lower voltage to support P/E cycles. Because of this, the cell can use a thinner oxide layer, while still reducing stress on the layer, resulting in higher endurance rates than drives with floating gate cells. The charge trap approach also enables faster read and write operations and lower energy consumption.
Charge trap cells have another advantage over floating gates. As floating gate cells become smaller, they also become more susceptible to disruptions, such as electrons inadvertently flowing from one floating gate to another. These disruptions can result in performance inconsistencies and lead to bad data. Because the charge trap layer is an insulator, such disruptions are less likely, and that makes it possible to shrink charge trap cells smaller than floating gate cells and produce denser drives with greater endurance.
Charge trap challenges
As promising as charge trap technologies sound, they do have challenges. According to the report "Reliability of 3D NAND Flash Memories," published in 3D Flash Memories, charge trap cells have several reliability issues. One of the most significant is data retention. Electrons can become trapped in the charge trap layer and start to accumulate, leading to data degradation, especially at high temperatures.
Floating gate cells offer better data retention because the charge in the storage layer is more stable, the report stated, and they result in fewer read errors and consequently fewer error-correcting operations. This is because a 3D flash drive that uses charge trap cells connects the nitride layers across all cells, providing a spreading path for the electrons.
However, the report also pointed out that floating gate cells are themselves susceptible to electrons coupling with those in other cells. In addition, charge trap cells can be made smaller and scale lower than floating gate cells, making them more attractive for hyperscaled arrays.
The evolving world of storage
Despite the challenges that charge trap cells present, they remain a promising technology for shrinking SSDs. It should be noted, however, that floating gate technologies have come a long way, too.
Jim Handy, an analyst at Objective Analysis, discusses the future of NAND flash.
Flash drives are denser, perform faster than ever and are getting cheaper all the time, substantially narrowing the gap between charge trap and floating gate drives. And new technologies such as 3D XPoint are changing the competitive landscape. Even so, drives based on charge trap architectures are likely to be around for some time and will improve with each new generation.