Intel dampened speed expectations--and the cool factor--of its Thunderbolt technology when it announced copper wire, not fiber optics, would carry data through the high-speed Thunderbolt ports. Still, 10 gigabit per second transfer speeds are nothing to complain about, and we can dream of the day Thunderbolt reaches its 100Gbps potential. By the time that happens, though, we may all be using something even faster.
A team of German researchers have used a technique called Fast Fourier transform to encode data into light and shoot it down a laser at 26 terabits per second. To put things in perspective, that’s 26,624 gigabits per second, or 2662.4 times faster than Thunderbolt’s current maximum speed.
The data is separated into more than 300 separate beams of light within the laser, then recompiled at the end to produce the correct information. Of course, that’s a gross simplification--the Fourier system actually puts the data back together based on the time each beam of light arrives at the reception point. By staggering the light beams to arrive in a certain sequence, the data is then easily recombined properly. The current setup at the Karlsruhe Institute of Technology in Germany sends the data over a 50 kilometer fiber optic cable before using the Fourier transformation to turn light back into readable data.
26 terabit transfers are possible through the beams of light carried within a single laser. And the optical technology scales with the amount of light added: an array of some 370 layers was able to transfer data at 100 terabits per second in an experiment. Amazing, but also prohibitively expensive and power hungry.
The current method does its best to replicate those speeds by splitting a single laser’s beam into a maximum of 325 colors, each of which can hold its own stream of data. A system employing fewer colors could still transfer data far faster than current technology--last year the team demonstrated a setup that moved along at a brisk 10 terabits per second.
The optical transformation system used to put the data back together could potentially be mapped to silicon, making it easier to commercialize. Copper will always be the more affordable choice, but there’s no speed like light speed. Will this be our next generation transfer technology?