EE Product News

Refresh! Flash Memory

By Mathew A. Dirjish, Associate Editor

Circa 1984 while in the employ of Toshiba, Dr. Fujio Masuoka invented a unique type of memory device having the desirable features of read-only memory (ROM), random access memory (RAM), and electrically erasable read-only memory (EEPROM, a.k.a., E2PROM) devices. Like RAM and EEPROM, one could write, erase, and rewrite data to the novel devices, or, akin to ROM, they could store retrievable static data for nearly infinite lengths of time. Also similar to RAM and ROM, Dr. Masuoka's memory had the propensity for large capacities. Importantly, comparable to ROM and EEPROM, the chips can store data indefinitely without the need for a backup battery or external power source (see Figure).

Aside from storing data without the aid of external power (non-volatile) and the ability to write and erase information, commonalities with EEPROM and ROM ended with the new memory's significantly faster read/write times and lower cost. Though not as fast as RAM, Dr. Masuoka's colleague, Shoji Ariizumi, likened the erasure speed of the new memory to that of a camera flash. The comparison stuck and flash memory was born.

Later that same year, flash memory debuted at the IEEE 1984 International Electron Devices Meeting in San Jose, CA. Foreseeing the benefits and economic viability of the technology, Intel introduced the first commercial NOR-type flash chip in 1988. Today, flash memory is found in a plethora of applications ranging from high-volume consumer products, i.e., digital audio players, cameras, mobile phones, USB drives, memory cards, and video games, as well complex designs such as embedded systems and integrated within microcontrollers.

Made up of gates, there are basically two types of flash memory devices: NOR and NAND. In operation, NOR flash performs like the typical RAM found in a computer, allowing direct access to a byte or bytes of space regardless of their position in the storage space. In use, NOR flash may be used to store a component's application-specific software, such as firmware in a router or a computer's BIOS. NOR flash specifies a working life of around 100,000 write cycles before developing bad blocks.

About a year after Intel unveiled the first NOR flash device, Toshiba developed NAND flash, which relies on flash-translation software that makes the device appear as a hard-disk drive to the operating system. This type of flash exhibits three distinct advantages over its NOR counterpart: a longer lifespan in the realm of one million read/write cycles, faster read/write times, and lower cost. In addition, NAND flash is capable of retaining larger chunks of data for either long- or short-term storage. The NAND devices are more useful for storing data collected by or downloaded to a product, i.e., information from a data logger, photos/video from a digital camera, music files on a MP3 player, and so forth.

A more recent and third type of flash is OneNAND flash. Created by Samsung, it supports faster data throughputs and higher densities, two major requirements for high-resolution photography, video, and other media applications. OneNAND could be viewed as a kind of hybrid of both NOR and NAND technologies. Essentially, a single OneNAND chip integrates a NOR flash interface, NAND-flash controller logic, a NAND-flash array, and as much as 5 KB of buffer RAM. In terms of speed, it can deliver sustained read rates up to 108 MB/s.

There are two types of OneNAND devices: muxed and demuxed. Address pins combine with data pins in the muxed type, while the demuxed chips keep both separate. When reducing pin count is a concern, the muxed OneNAND may be the way to go. Additionally, muxed OneNAND operates from 1.8V exclusively and demuxed, with its lower density of less than 1 Gbit, offers both 1.8V and 3.3V options. If the density of either a muxed or demuxed device exceeds 1 Gbit, 1.8V operation is the only option.

In summary, NOR flash is suitable for code storage, that is, firmware, device application, etc., and NAND flash handles mass storage chores similar to a hard-disk drive. Providing the best of both worlds, OneNAND flash is competent for both code and mass-data storage as well as being more power efficient.

Each type of flash memory is comprised of a number cells with each cell being an array of floating-gate transistors. Referring to early flash devices, each cell stores one bit of information. However, continual development has led to multi-level cell devices that exceed the one-bit per cell by using more than two levels of charge.

A typical NOR cell appears like a MOSFET with two gates: a control gate and a floating gate with a layer of oxide that insulates it from the control gate. Sitting between the substrate and the control gate, the floating gate is where data is stored. Programming a NOR cell occurs when current flows from source to drain, which incurs a high voltage on the control gate. When this voltage reaches the proper level, electrons (data) flow into the floating gate where they are retained with the help of the insulating layer. This process has been dubbed hot electron injection.

Via a technique called tunnel release, erasing a cell, resetting data to all ones, requires the injection of a large voltage differential between the source and the control gate. This voltage differential, delivered via an integrated charge pump, forces electrons off of the floating gate, thereby erasing the cell.

Operating from 3.3V or 5V supplies, NAND flash chips employ the same tunnel-release technique for erasing data as NOR devices. For writing, tunnel injection, a quantum tunneling effect also known as Fowler-Nordheim tunnel injection, is a method whereby charge carriers are injected into the floating gate through the oxide insulator. This approach is said to save power while decreasing write times. A side-by-side comparison of NAND and OneNAND flash performance parameters is shown in the Table.

Probably the most critical limitation to flash memory is the finite number of write/erase cycles. Most commercial flash-based products are guaranteed to endure up to one million write cycles. This number may seem like a lot and, in the case of NOR flash, it probably is, since storing an application or BIOS for a very long period of time may present no concerns. However in typical NAND applications where files are written, retrieved, and overwritten regularly, cycles get eaten up pretty fast and most users will probably not keep count. For on-going storage of critical data where updating is frequent, flash may not be an option.

To counteract this limitation, firmware or file-system drivers can be employed that literally count the number times the memory is written to. These applications will dynamically remap blocks so that write operations are shared among the sectors. Alternatively, in the case of write failure, the application, via write verification and remapping, delegates write operations to unused sectors.

Like RAM, flash memory can be read or programmed a byte or a word at a time, but erasure must be a complete block at a once, which resets all bits in a block back to one. This translates into more time required for reprogramming. For example, if one bit (0) is written to a block, in order to reprogram the block it must be completely erased as opposed to merely overwriting the bit.

The advantages of flash memory far outweigh the limitations, as evidenced by the array of flash-based products on the market and their popularity. Like legacy memory and storage devices, flash capacity started off small and is now approaching the level of hard-disk drives found in notebook computers. For example, around this time last year, Samsung introduced a 32-GB flash drive based on its 32-MB flash chip (K9F5608U0BYCB0) using NAND technology and, shortly before that, 32-Gb chips, based on a 40nm process. Another example on the hard-drive end is BiTMicro's Edisk, which provides 155 GB of storage capacity on a 3.5" solid-state disk.

Memory cards, the favored storage medium for digital cameras and other portable products, are now comparable in capacity to what were considered state-of-the-art hard drives of the late 1990s. Currently, 1 GB and 2 GB capacities have supplanted the 256MB and lower capacity cards. The same applies to USB flash drives, also called thumb drives and/or pocket drives.

Overall, there are five obvious benefits associated with flash memory: it is small and will most likely shrink further over time, low power, high capacity and will expand exponentially over time, nearly indestructible (no moving parts) when packaged properly, and it is becoming less expensive.

Obviously, flash memory is one of those technologies that will be with us for some time to come. New storage technologies that will be derived from it are also on the horizon.

Just three months ago, the MultiMediaCard Association (MMCA) and the JEDEC Solid State Technology Association agreed upon eMMC as the trademark and product category of a class of embedded memory module products for embedded flash memory applications built on the joint MMCA/JEDEC MMC specification. The eMMC label details an architecture made up of an embedded storage system employing a MMC interface, flash memory and a controller, all of which are housed in a small BGA package. The architecture is expected to win favor in numerous products including industrial, mobile phones, navigation systems, media players, and other portable electronics devices.

The system, based on MMC System Specification v4.1/4.2 and JEDEC BGA packaging standards, specifies interface speeds up to 52 MB/s. Importantly, the standard overcomes the operating-voltage limitations of some current flash memories in that it supports an interface voltage of either 1.8V or 3.3V.

With eMMC, the host system can access all mass-storage memory including memory cards and hard disk drives via one MMC interface protocol bus. This system architecture is far more flexible than that based upon other memory card-only standards. In turn, the standardized eMMC protocol interface keeps complexities, like NAND flash functional differences, invisible to the host. Additionally, because eMMC is an industry standard, multiple sources for memory components will be available.

Also gaining momentum, QUALCOMM has announced its decision to adopt Samsung's OneNAND flash memory for use in all forthcoming Mobile Station Modem chipsets. Support for the memory, considered to exhibit the fastest read/write speeds available, already exists on a few chipsets, but will be the norm for a wider range of emerging multimedia products.

A prime target for the OneNAND-based chipsets will be 3G phones. The memory, with its 17-MB/s write speed, can offer guaranteed continuously streaming air downloads beyond the HSDPA specification. In addition to multimedia handset designs, OneNAND promises to be a valuable option for use as a non-volatile buffer in hybrid hard disks. Currently, the chips provide capacities up to 2 Gb using a 60nm process technology, and that will likely expand to 4 Gb at 50-nm later this year.

Flash Memory Web Resources


JEDEC Solid State Technology Association

MultiMediaCard Association

Open NAND Flash Interface Working Group



How Flash Memory Works

AMIC Technology Corp.

Atmel Corp.

Eon Silicon Solution Inc.

Excel Semiconductor Inc.

Micron Technology, Inc.,

Msystems Inc.

SanDisk Corp


Winbond Electronics Corp.


Product URL: Click here for more information

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.