eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
Researchers from University of Massachusetts Amherst and Texas A&M University have succeeded in writing information to flash memory under low-voltage conditions, paving the way for a new generation of low-power gadgets that can store data. They presented the paper Feb. 16 at the USENIX File and Storage Technologies Conference in San Jose, Calif.
Flash memory generally requires 2.2 to 4.5 volts, which makes them unusable in devices using low-power microprocessors, the researchers wrote. For example, the MSP430 microcontroller from Texas Instruments is intended for embedded applications and runs on as little as 1.8 volts.
There are a number of memory manufacturers that have started building low-power flash solid-state-drives for use in embedded storage systems, such as Greenliant and SanDisk. Intel also announced last month a SSD that can be used in industrial embedded applications. Greenliant targets embedded devices in enterprise, industrial, automotive and networking applications.
In general, designers have opted to either boost CPUs to meet flash memory’s minimum voltage requirements, or just not used flash memory at all, the researchers said. Those gadgets would not be able to store data.
Tablets and netbooks can support flash memory because it has the room to support separate power rails for the CPU and flash memory. That is not possible on small gadgets where the flash memory is integrated within a microcontroller. If the battery can support both, that’s fine, but generally the batteries are also much smaller in size and power.
The most trick to effectively write data to flash memory when running on less than the minimum required voltage was “persistence,” according to the project’s lead researcher, Mastooreh Salahegheh. The software-only coding algorithms exploited “the electrically cumulative nature” of half-written data based on a quantum mechanical phenomenon called tunneling. In this case, outside electrons travel to the chip a little at a time and accumulate since it’s not being used. Once enough electrons have been collected, there’s enough power to meet specifications to write data for that instant, the researchers said.
Obviously, this process won’t have good performance or efficiency, but if the goal is to conserve power, it accomplishes it readily, according to the paper. On a sensor-monitoring application using the MSP430, researchers used persistence to reduce overall energy consumption by 34 percent, according to the paper.
“Our evaluation shows that tightly maintaining the digital abstraction for storage in embedded flash memory comes at a significant cost to energy consumption with minimal gain in reliability,” according to the paper.
A persistence method like this may be useful for a number of small devices, such as remote control key fobs or digital picture frames. It can also be useful on devices that currently don’t have batteries, such as RFID tags and electronic passports. With the persistence technique uncovered by Salajegheh’s team, embedded application designers can fit in non-volatile storage inside these small devices.
Salajegheh, Kevin Fu, and Erik Learned-Miller were the researchers from University of Massachusetts Amherst. Yue Wang and Anxiao Jiang from Texas A&M made up the remainder of the research team.