American Scientist

Several recent studies have highlighted the fact that our central nervous systems often take advantage of "stochastic resonance," a curious phenomenon that arises when a signal that would normally fall below the threshold of perception can nevertheless be detected by virtue of added noise. That the presence of random noise can help a weak signal to be registered is admittedly counterintuitive. But research has clearly shown this to be the case for many physical and biological systems.

Some investigators are now applying this knowledge clinically. One group has, for example, intentionally added noise to the signals generated by cochlear implants so as to mimic the random firing of the auditory nerve; this lowers the threshold at which patients fitted with these devices can detect subtle sounds. Another group has recently found a way to use added noise to enhance balance control in older people whose feet no longer sense the ground well.

It turns out that the phenomenon of stochastic resonance can help electronic brains too, allowing computers to become more sensitive to subtle signals in their networked environment. A demonstration comes from yet-to-be-published work of Joel Lepak, an undergraduate math major at Youngstown State University in Ohio, and Michael Crescimanno, a faculty member in the Department of Physics and Astronomy at the same institution.

Lepak and Crescimanno devised a way to measure the transmission time between interconnected computers with astonishing precision. Indeed, they were able to do so with a fine enough resolution to gauge the speed at which data packets bounce back and forth through a few tens of meters of standard computer cable. To do this they ran ping, a program that network administrators often invoke to test connections between computers. Because ping can only resolve transmission times to the nearest microsecond, it is not obvious how it can be used to clock the passage of signals traveling through such a short stretch of cable at a good fraction of the speed of light. The total back-and-forth in such a setup takes no more than a few hundred nanoseconds, which is below the nominal threshold of detection.

"Millions of people have seen ping logs," Crescimanno notes. But it seems that only a small number have ever taken the time to ponder the fact that the return time between two computers is not always the same. Seeing that, Crescimanno, who had read about stochastic resonance in the biophysics literature, reasoned that this phenomenon might allow him to improve what one can achieve with ping. The noise in this case is inherent to the network itself—computer clocks running out of sync, radio-frequency signals picked up through network cables and power lines, and so forth. These sources add some jitter to the return delays registered when a computer is programmed to ping another a large number of times. Plotting the average travel time as a function of the length of cable connecting two computers reveals a line whose slope is the signal velocity—in this case about two-thirds the speed of light.

So long as the network connection is not too noisy, a good result appears easy to get. "Basically, it worked the first time," says Crescimanno, who is promoting this procedure as an exercise for undergraduate physics labs. He has been using a similar strategy to offer students a way to estimate the size of the Earth: Instead of pinging, he uses traceroute (another widely available network-diagnostic tool) to send data packets between two computers that are connected on the Internet through a long transoceanic cable. This program shows the signal path and the time required for packets to make a round-trip journey to the various computers encountered along the way. After measuring the path of the subsea cable on a globe, students can work out the planet's circumference from the speed of the signal and the amount of time it takes to make the large transoceanic hop.

Crescimanno hopes eventually to be able to detect the so-called Sagnac effect using much the same approach. This effect, first demonstrated in the early part of the last century, arises when one compares the travel times of a signal sent clockwise around a loop with one that goes counterclockwise. If the path length is large enough (or the sensitivity of the apparatus fine enough), one can measure a difference in travel time when the experimental platform is rotating. Crescimanno intends to measure the slight difference in travel times between data packets as they make their way around a large triangular segment of the Internet that has one node on each of three continents. He figures that with the resolution achieveable through stochastic resonance (which can be as fine as 50 nanoseconds, or about 1/20th the nominal resolution of ping), he may well be able to detect the Earth's rotation, which with an optimal geometry would produce a Sagnac effect that is about 500 nanoseconds. Whether or not Crescimanno achieves this ambitious goal, one cannot help but admire what he and Lepak have already demonstrated: that computers and the Internet can be used to teach physics without having to wow students with flamboyant displays of Java, Flash, Shockwave or Quicktime—indeed, without having to produce any fancy graphics at all.