When it comes to signal processing, timing is everything.
IEEE Senior Member Yonina Eldar and her team of graduate students at Technion-Israel Institute of Technology, in Haifa, have found a faster, more efficient way to process data transmitted by wideband signals. They built a prototype that can interpret the signals used in applications such as radar systems and medical imaging by taking a smaller than normal number of data samples while still producing high-resolution images. Their modulated wideband converter uses less power and processes signals faster than existing sampling equipment, which is often expensive and unwieldy. She is working on several applications of the technology, including improving radar image resolution, coginitive radio, and making ultrasound machines process images faster.
Eldar, a professor of electrical engineering at Technion in Israel, was awarded the 2011 Weizmann Prize for Exact Sciences by the city of Tel Aviv for her work on the converter and on sub-Nyquist sampling. The award, named for Chaim Weizmann, the first president of Israel, is given to groundbreaking original research in exact sciences.
FROM THEORY TO PRACTICE
Eldar's converter stemmed from mathematical curiosity, she says. She wanted to know whether there was a way around the Nyquist sampling theorem, which states that to digitally process a continuous-time signal, it must be sampled at a rate that is twice the maximum frequency. That means when converting analog signals to digital, numerous samples must be taken over an extended period of time. Even more samples must be taken with ultrawide-band signals. The theorem forms the basis for how most of today's digital converters work.
"The Nyquist theory is kind of a worst-case scenario—it assumes that you know nothing about your signal's structure other than its bandwidth," Eldar says. "But in modern applications, like radar or channel estimation, there's a lot of structure to signals that is used in various signal processing algorithms. So we wanted to figure out whether we could use that structure to reduce the sampling rate and still get high-resolution images in the end."
The team theorized that by using what one already knows about a signal's structure and identifying gaps in the spectrum or in time, fewer samples could be taken—thus less time is needed to process them—without damaging the signal.
Initially, the team had no intention of proving the theory in the lab. "When I talked to colleagues about our theory, especially in industry, they gave me a funny look, saying that it seemed like an interesting idea but it was never going to work," she says. "I started to think they were right, so we built the first prototype just to learn where we might have gone wrong. We didn't think it would actually work."
The prototype circuit board samples analog signals at a very low frequency. Those samples were then digitized, transmitted to a computer, and processed and analyzed using MATLAB simulation software. Later, with the Technion's High Speed Digital Lab and National Instruments, the entire processing chain was performed on a FPGA. "We did a lot of hard work in the lab," Eldar says, "but we didn't have to change or rewrite any part of our theory to prove it out."
AN ARRAY OF APPLICATIONS
Eldar and her team are focusing on numerous uses for the converter, including wireless communications, defense, and health-care applications. "I have a strong interest in the medical field," Eldar says. "I like knowing that the work I do can potentially help people who need it most."
She and her graduate students are working with General Electric Israel on ways to downsize today's large ultrasound machines into portable, even handheld devices. The prototype they're working on applies their sampling technology to produce clearer ultrasound images in a fraction of the time.
"By using our method, each time the machine takes a data sample, you get more information per data point, so you need fewer snapshots," Eldar explains.
She is also working with defense contractors to improve the image resolution of radar systems and to enable cognitive radio transmission and reception. In the future, she is considering applying her sampling technology to vehicles. By decreasing the time it takes to process signals transmitted by a car's sensors, the vehicle can react more quickly to dangerous situations.
"That's what makes engineering fun—I get to learn about all of these different applications and hopefully be able contribute to a variety of areas," she says.
A LEARNING PROCESS
Eldar developed a passion for signal processing during her time as an undergraduate student at Tel-Aviv University, where she earned bachelor's degrees in electrical engineering and physics. "I had two amazing signal processing instructors—Prof. Ehud Weinstein and Prof. Arie Yeredor," she says. "It made me realize that good teachers are such an important factor in determining what you wind up doing."
She went on to receive a master's degree in electrical engineering from Tel-Aviv in 1996. After earning her Ph.D. in electrical engineering and computer science at MIT in 2002, she returned to Israel, where she is now a professor of electrical engineering at the Technion. She worked primarily with signal processing theory and coding. Until she began working on beating the Nyquist theorem three years ago, she had never actually built hardware.
"Now my lab is very active," she says. "We still do a lot of theory, but we are very involved in building prototypes, so even if students are doing purely theoretical work, they're connected to what's going on in the lab.
"I'm extremely excited about this line of research, and I hope the converter will be used in future devices. But even if it doesn't take off, I really learned lot from the process, especially the importance of connecting theory and practice," she says. "Puting in the effort to actually build prototypes and move from the paper to the lab is an exciting and crucial step toward impacting future technologies."