UCSB has always been a friend of the Internet.
The service started here in the late 1970s as DARPAnet, the Defense Dept.’s communications network. Though the Net has gone public, DARPA, the Defense Advanced Research Projects Agency, is still funding cutting edge computer engineering research today. Five UCSB researchers were recently awarded $3.5 million by DARPA to create advanced routing devices for optical computer networks over the next four years. The project is part of a field of research known as photonics.
Because of their ability to transmit large amounts of data quickly, fiber optic networks are a popular solution among engineers for creating high-speed computer networks. A fiber optic network transmits data using light, rather than electrical signals.
Currently it is much easier to steer an electrical signal through a network than an optical signal. The act of steering a signal is known as routing. Currently, in order to route an optical signal, the signal must be changed back into electrical information, assigned a direction, and then converted back into an optical signal again. This interplay between electrical and optical signals is known as optoelectronics.
Routing a signal using optoelectronics takes a substantial amount of power. One of the goals of photonics is to eliminate optoelectronic devices as the primary routing devices for optical networks. UCSB researchers’ best hope for achieving this goal lies in a three-centimeter chip manufactured at UCSB’s engineering complex.
“That’s huge,” lead researcher Dan Blumethal said. “The long-term plan is to push that density up.”
Despite its bulky size, the chip is an elegant piece of engineering. It employs a strategy more familiar to astronomers than computer engineers: optical interferometry. In optical interferometry, light from two sources is combined into a single signal. The two light sources interfere with one another. This interference may be constructive or destructive – that is to say, the combined signals may reinforce one another or cancel each other out.
In the routing chip, a colored laser is fired into a beam splitter, which splits its signal into two independent signals. These signals then travel through two passageways known as modulators. The modulators amplify the two signals, and pass them on their way. When the signals meet again after their trip through the modulators, they interfere with one another. If the two paths were exactly the same length, the wavefronts of both light signals will still be in synch with one another and the light will pass on as if nothing happened. If, however the pathways were of different lengths, the two signals will not match up correctly and will cancel one another out.
The modulators are set up so that one pathway is normally longer than the other. Because of this, the laser continuously fires, but no signal gets through. The chip is designed so that when a pulse of light is received from the outside network, the light from that signal goes through one of the modulators and actually changes the index of refraction of the modulator. The index of refraction is the speed of light inside a material. By changing the speed at which light travels through the modulator, the apparent length of the modulator is changed.
When the length of the modulator is altered, the lengths of the two pathways of light match up and the signal from the laser is allowed to pass. Meanwhile, the original signal from the network is siphoned off. The laser is always a different color from the light making up the original network signal.
The effect of all this is that a signal enters the chip as one color and leaves as another. The laser is a variable laser, so the user of the network can pick the color of the light leaving the chip. All these functions have never before been integrated into a single chip.
The first photonic networks may be set up like old phone systems, in which a circuit is formed between a sender and a receiver. The connection will be formed by assigning a color of light. If a signal is sent in green light, users may pick up the signal by tuning into green light, much like one would tune a radio. If all green circuits are “busy,” the routing chip could change the light to blue and effectively open up a new line.
The final goal of such systems, however, is to produce a network that works like today’s electronic network in which outgoing data is divided into small packets of information, each is assigned a destination, and routed separately. Different colors of light pass at different angles through a prism. By controlling the color of the light information is transmitted in, devices like those produced at UCSB may eventually be able to route packets of information at different angles through a prism to produce a routing system much like those used today in electronic networks. Such systems will have to wait until the chip can be scaled down to a manageable size, though.
What’s more, a light signal does not necessarily need to be a digital signal. The amplitude of light has long been used to convey analog signals – AM radio for instance. The DARPA grant is concerned with creating networks to route both analog and digital signals. Preserving of the integrity of an analog signal over a network is a difficult problem – one worth about $3.5 million.
The grant will be investigated by Dan Blumenthal, John Bowers, Larry Coldren, Evelyn Hu and Nadir Dagli.