Information Sciences Institute

Sending data happens quickly. As quickly as the speed of light to be precise. But processing data, or compressing it, is a much slower, energy-consuming process that physicists have struggled to navigate; that is, up until now.

Typically, data is only in the form of “light” when it is to be transmitted from one point to another, and it travels at blazingly fast speeds.

“Whenever you need to send a lot of data fast, it’s in the form of light,” explained Jonathan Habif, an experimental physicist, USC Viterbi research assistant professor of electrical and computer engineering and research lead at the USC Information Sciences Institute (ISI)’s Boston office.

But processing the data is an entirely distinct step.

“Whenever you have to process data, it’s done through electrons inside a computer. To move from communications to processing, you have to move from light to electrons back to light.”
It requires transmitting the data to a computer where it can be converted into electrons. Converting and processing data into different energy forms is a much longer operation relative to data transmission. In fact, data processing is generally limited to less than about 1% of the speed of data communications.

A novel optical processing technology

Supported by the Defense Advanced Research Projects Agency (DARPA), engineers and physicists at USC Viterbi’s Information Sciences Institute (ISI) have created a novel technology that consolidates the time-consuming nature of data processing and reduces the amount of energy required from start to finish.

Habif’s research team proposed a technology that would consolidate data transmission and data processing. The team demonstrated an “optical correlator”, which enables them to begin processing data even before light is converted into electrons. Doing this keeps the data signaling and processing at the speed of light.

“We are building light circuits that can perform a certain amount of processing on the data as it’s flowing through the channel,” said Habif. By initiating the step of data processing while it is still in the form of light, it increases efficiency along two vectors: the technology decreases latency, which typically clogs up networks, and it also consumes less power.

Another facet to this technology includes its ability to monitor the flow of data between different communication channels. The optical correlator looks for a particular pattern of data flowing through a channel, generating an alert if an abnormal pattern is detected. This could potentially mitigate the possibility of unwanted personal information from flowing into other communication channels.

Providing power without the power

Alan Willner, Distinguished Professor of Electrical and Computer Engineering at USC Viterbi, collaborated with Habif to devise this technology and explained its greater implications in advancing how circuits can utilize power, even when an external power supply is not directly present.

Not only does the new correlator enable greater efficiency in data processing, but the techniques implemented also eliminate the need for an external power supply to the circuits — namely, electrical power.

“Normally, if you have optical circuits or an optical processing block, you still have to apply electrical signals,” continued Habif. This discovery distinctly changes the way data signaling and processing can be conducted.

If a power outage were to occur at a given location, the technology offers the flexibility to reinstate power and to control the functionality of that given node from a remote routing node.
The power comes from a transmitting node, which is transferred as light to the remote routing node.

“You’re throwing an optical signal from 10 kilometers away and turning it into a voltage supply for the switch,” said Willner, the Steven and Kathryn Sample Chair in Engineering , explaining the mechanism of transferring power from a transmitting node. “If a switch doesn’t have any local power, we’re sending power to it,” continued Willner.
The light is then converted into power that’s used to control the switch that was initially disabled from a power surge — all at a remote distance.

In addition to finding novel ways for controlling remote nodes, another innovation to their project includes monitoring the switches to evaluate its effectiveness in powering other nodes at a distance.

“Imagine you have a train that’s going to L.A. and there’s a fork in the track,” said Willner, “with one veering off to the left to L..A and the other veering off to the right to go to Denver. Somebody’s there to switch the train tracks with a lever to control the direction of the train.”

But the question they’re asking is what if nobody is there to pull the lever? “We’re sending that signal and the power to do it, and we’re monitoring it to make sure it goes in the right direction, even if there’s no person there,” explained Willner. The technology ensures that the data, like the train, is going in the right direction.

Another exploration for the future is looking at the maximum power that can be transmitted without any battery or external source, and just how quickly this can be done. Habif explained that silicon photonics (circuits that use light rather than electrons) has the potential to power optical circuits faster and allow a user to implement any data processing function they would want. Professor Hossein Hashemi from the USC ECE department is leading the team in moving this technology into silicon photonics.

The all-optical correlator technology developed by the team represents just a first step in understanding the limits to data processing in the optical domain. As the team moves forward under DARPA sponsorship they plan to continue pushing the boundaries between optical communications and data processing in pursuit of a communications network architecture where engineers and users are unconstrained in a data communications environment enabling a new set of applications in fields as diverse as high-performance cluster computing and information security.

Published on June 14th, 2022

Last updated on June 14th, 2022