Studying with the radio on may not be the best way to remember what you’ve read. But scientists have now built a data storage device whose memory gets a boost from noise.
The device can store one bit of information, such as a 0 or a 1, only when surrounded by electronic noise, which is normally a problem in computer circuits. The experiment was published November 4 on ArXiv.org.
“The whole thing works very well with some amount of noise,” says Diego Grosz of the Instituto Tecnológico de Buenos Aires, a coauthor of the study. “If you remove the noise, it doesn’t store the bit at all.”
Noise in the form of stray electrons and heat plagues computer circuits. As transistors get smaller, noise becomes more of a problem, and the computer’s error rate goes up.
To get around this problem, some engineers are trying to take advantage of a phenomenon called stochastic resonance, in which a system uses noise to perform better. The effect has long been observed in neural systems. Earlier studies have suggested that systems using stochastic resonance could encode basic logic operations, or even embed information in the noise itself.
“Noise is everywhere, you cannot avoid having noise,” Grosz says. “We thought, ‘How can you put that to good use?’ ”
Grosz and his colleagues built a system of two oscillators that can hold a harmonic wave at a specific voltage representing a 1 in the logical language of computers. Without noise, the wave disappeared quickly. But with noise, it persisted at that voltage for a long time, even when the power source driving the wave was turned off.
The right amount of noise helped hold the oscillators at their peak performance level, similar to how pushing a swing at the right time makes it go higher and higher.
The researchers found that there was an optimum amount of noise that made the wave last longest at the right voltage. Any other voltage was considered an error.
Though the system Grosz and colleagues built for this experiment is about 12 centimeters long (about the size of a reporter’s notebook), Grosz says in principle it could be made small enough to be used in computers.
“You should be able to do this with very small transistors, the ones you have in CPUs,” he says. “We didn’t do it, but it can be done.”
Laszlo Kish of Texas A&M University says that “it’s an interesting effect, but it’s not for practical use.” He points out that even though the error rate in the new device is lower, it doesn’t carry as much information as a noiseless circuit would.
“It’s like how the food industry adds chemicals to make food keep for a long time,” he says. There’s an optimum amount of chemicals for preserving food, just like there’s an optimum amount of noise for stochastic resonance. “But fresh food is still the best.”
Grosz isn’t certain his circuit is ready for prime time, either. “Is this thing going to see the light of day? Is this going to be inside a computer at some point? I surely hope so, but I don’t know,” he says. “We’ll keep trying.”
