by October 1, 2001 0 comments

From the humble origins of your PC’s BIOS or your notebook’s PCMCIA card, flash memory is today one of the hottest development areas. This is largely driven by the ubiquitousness of gadgets like cellphones, PDAs, digital cameras, and MP3 players, all of which use some form of flash memory to store data.

Flash memory is an EEPROM (Electronically Erasable Programmable Read Only Memory) chip, and though it is read-only memory, you can electronically erase and write data on it. Flash memory offers the advantage of solid-state storage, which means that unlike RAM and like the BIOS chip in your PC, the data you store on it doesn’t disappear when you switch off power. Its also faster, smaller, and lighter than conventional storage media like hard disks. Also, unlike other EEPROM chips that have to be erased and programmed one byte at a time, flash memory is programmed in blocks, that is, several bytes at a time, which makes it faster to use.

Inside a flash-memory cell

Like other memory chips, a flash-memory chip is a silicon wafer on which millions of flash-memory cells and other circuitry are etched. The cells are etched in an array of columns (called bitlines) and rows (called wordlines), and their intersection is called the address of a memory cell. A cell stores one bit of data, but recent developments have also thrown up what is called a multilevel design, which lets a flash-memory cell store more than one bit. We’ll talk about this a little later.

Each flash memory cell has a specific threshold voltage and is made up of two transistors–the control gate and the floating gate–and source and drain regions of the transistor. The floating gate is surrounded by an insulating oxide, and is connected to the wordline only through the control gate. If this connection is in place, the memory cell has a value of 1 and conducts during a read operation. If this connection is not in place, the cell has a value of 0 and does not conduct during a read operation. The charge on the floating gate determines whether the cell will have a value of 0 or 1. A method called Fowler-Nordheim tunneling is used to erase the data on these cells by transferring electrons from the floating gate to the insulating oxide layer using a high-voltage charge. Another one called CHE (Channel Hot Electron) injection is used to program them by transferring electrons from the insulating oxide layer to the floating gate.

Types of flash memory

Depending on the way memory cells are connected to the bitline, there are different kinds of flash memory, suitable for different purposes. Two most commonly-used types are NOR and

NOR is characterized by parallel connection of the memory cells to the bitline. Here, if any of the cells in a bitline turns on (has a value of 1), the whole bitline goes to 0. This architecture is good for random access, and has fast read speeds, but relatively lower erase and write speeds. So, it is mostly used to store programs for execution in devices like

NAND has a series connection of memory cells to the bitline. Here, if all the memory cells in a bitline turn on, the bitline goes to 0. NAND is characterized by slow read speeds, but faster erase and write speeds. So, it’s most commonly used for solid-state storage in devices like digital cameras or MP3 players. Also, NAND flash’s series connection requires less area and hence enables more number of bits to be stored on a smaller silicon chip.

Future directions

Developments in flash memory have focused on the inter-related areas of reducing price, reducing the size of the chip without affecting performance.

Apart from market conditions of demand and supply, prices of flash memory also depend on cost of the chip, which depends on die area and die yield. Reducing this size will bring down costs and hence prices. The MLC (multi-level cell) design is an important development in this regard.

In a conventional flash-memory design, each cell stores only one bit of data, either 0 or 1, which are differentiated by the amount of charge on the floating gate. If this charge is above a certain threshold voltage, the cell stores a 0, and if it’s below this voltage, it stores 1. MLC design works on the principle of storing more than one bit of data on a cell by enabling each cell to distinguish more than two charge states. That is, if four charge states can be distinguished, two bits of data can be stored on a cell; if eight charge states can be distinguished, three bits of data can be stored, and so on. This helps in packing more data on a chip, without increasing the die area or reducing the die area for a given amount of data storage capacity. Intel’s Strata

Flash is an example of MLC flash.

Improvements in the structure of the cell to increase its life have also happened. When flash memory is erased and reprogrammed repeatedly, the transfer of electrons from and to the insulating oxide layer of cells damages the oxides, leading to lower data retention or other problems. SST (Silicon Storage Technology) has patented a cell structure called the thick-oxide split gate cell, where the thickness of the insulating oxide layer–about 40 nm compared to 10 nm for other cell structures–helps to minimize these problems. Another factor that will lead to improved performance is reducing the voltages at which read and program operations take place on a flash-memory device.

An interesting fact about flash memory is that unlike RAM, where the architecture of any type of RAM will be standard irrespective of the manufacturer, there is a proliferation of flash-memory devices–Memory Stick from Sony, Secure Digital from Panasonic, SmartMedia (originally developed by Toshiba), CompactFlash (originally developed by SanDisk)–each of which have their own architecture to deal with the issues mentioned above. These are also not interchangeable in use in that if your digital camera supports one of these, you can’t use another one in its place. For this, the manufacturer of your digital camera has to build in support for more than one device. The use of flash memory is growing today and there is enough space for more than one standard, but it will be interesting to see how the market dynamics change as the industry begins to hit a ceiling.

Pragya Madan

No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.