Study details how analog computing solves signal processing challenges faster

Researchers at RIT have improved the electronics used in communication and radar systems to better process signals using electromagnetic radio waves. This breakthrough could advance computing processes for technologies that need to handle signals quickly and efficiently.

Mohammad-Ali Miri, an associate professor of electrical engineering in RIT’s Kate Gleason College of Engineering, worked with international partners to improve the use of different frequency ranges of electromagnetic waves. Their work, described in a fall issue of Nature Communication, alleviates one of the key areas where bottlenecks have occurred, specifically in the matrix operations where computing processes occur.

Matrix operations comprise complex sets of mathematical operations used to perform computing tasks. These operations are conducted by handling structured data grids of multiple variables simultaneously, allowing for the identification of data patterns, modeling, and translation.

Miri was one of the project leaders contributing to “Programmable circuits for analog matrix computations.” The international team of engineering researchers is led by Rasool Keshavarz, professor and senior research fellow at the University of Sydney; Kevin Zelaya, post-doctoral fellow in RIT’s Department of Electrical and Microelectronic Engineering; and Negin Shariati, associate professor at the University of Sydney.

The team has been working on improving the structures that can do matrix operations on electromagnetic waves and implementing them in photonic systems. The physics behind photonic systems follows a mathematical scheme analog to that of classical computing, and the team looked to apply these theories to radio or microwave frequency, Miri explained.

“We know the physics may be the same, but the technology is completely different when you go to a different frequency range and it is a whole different design and fabrication process and even characterization,” he said. “We are in a race for advanced chips, and the artificial intelligence that makes certain industries or companies very popular is that they can do matrix operations very fast and at low power. But the thing we know with digital computing, we are reaching the limits with those approaches.”

Miri and the team sought ways to improve how signals come into, and are analyzed by, radar systems. These systems are energy-demanding and solutions to decreasing those demands while retaining reliability and providing faster analysis were being assessed. The array of antennas in radar systems reflects signals and gathers data information to be processed using matrix operations. The electromechanical waves through space are detected by the antennas, then converted to digital signals through an analog-to-digital converter.

By developing a new type of microwave circuit that directly processes radio signals by carefully controlling how waves interact, the device bypasses any digital bottleneck, therefore performing complex operations at the speed of light. The team has also shown that this method achieves comparable performance at a lower power than a conventional processor.

“That is huge pre-processing that can significantly reduce the load in the next stages of the whole system,” Miri said. “This is how we begin to encode information in this system. The way the device is doing the operation is that physics of the system. That is the new and interesting thing with this device. Its architecture is what we have developed.”