Professors Qiangfei Xia and Joshua Yang of the Electrical and Computer Engineering (ECE) Department at UMass Amherst have just published an article about their research on a “computing engine using large memristor crossbars” in the opening issue of Nature Electronics, a research journal launched recently by the Nature Publishing Group. Xia, Yang, and their research colleagues say that “Memristor crossbars offer reconfigurable non-volatile resistance states and could remove the speed and energy efficiency bottleneck in vector-matrix multiplication, a core computing task in signal and image processing.”
As Xia explains about the significance of their research, “We live in an analog world, meaning the information we receive is continuous in time without breaks. For example, music you hear in a concert is a continuous sound wave, an analog signal. But the CD you buy in the theater foyer carries the music that has been converted into a discrete digital signal. When it is played from your stereo, the digital signal is converted back to an analog one. These conversion processes are power hungry and susceptible to noise. With the memristor crossbar we built, we can process analog signals directly without those conversions.”
Xia adds that “The large memristor array can do computation at a much higher rate. It is like counting beans, you count one by one in a digital processor. But now with the memristor crossbar, you can count all, thousands or millions or more, at once.”
“In addition,” says Yang, “the memristor crossbar does in-memory computing, meaning the same devices perform both information storage and processing, which obviates the need for shuffling information and thus breaks the biggest bottleneck of our current computing architecture.”
The co-authors who collaborated with Xia and Yang on the Nature Electronics article, entitled “Analogue signal and image processing with large memristor crossbars,” are: Can Li, Yunning Li, Hao Jiang, Wenhao Song, Peng Lin, and Zhongrui Wang of our UMass ECE department; Miao Hu, Eric Montgomery, Noraica Dávila, Catherine E. Graves, Zhiyong Li, John Paul Strachan, Jiaming Zhang, Ning Ge, and Stanley Williams of Hewlett Packard Labs in Palo Alto, California; and Mark Barnell and Qing Wu of the Air Force Research Laboratory Information Directorate in Rome, New York.
According to the Nature Electronics article, using memristor crossbars to multiply an analog-voltage-amplitude vector by an analog-conductance matrix at a reasonably large scale has proven challenging due to difficulties in device engineering and array integration.
“Here we experimentally demonstrate that reconfigurable memristor crossbars, composed of hafnium oxide memristors on top of metal-oxide-semiconductor transistors, are capable of analog vector-matrix multiplication with array sizes of 128×64 cells,” as the researchers say in their Nature Electronics article.
The researchers note that their output precision (five to eight bits, depending on the array size) is the result of high device yield (99.8%) and the multilevel, stable states of the memristors; while the linear device current-voltage characteristics and low wire resistance between cells lead to high accuracy. “The current work realizes such an analog computing engine at a large scale and with a high accuracy for the first time,” Yang adds.
“With the large memristor crossbars,” as the researchers explain in their article, “we demonstrate signal processing, image compression, and convolutional filtering, which are expected to be important applications in the development of the internet of things and edge computing.”
Professor Xia received his Ph.D. degree from Princeton University in 2007. He joined the ECE faculty at UMass in 2010, after spending three years as a postdoctoral research associate at Hewlett Packard Labs. Among his honors, he has received an NSF CAREER Award, a Defense Advanced Research Projects Agency (DARPA) Young Faculty Award, the Barbara H. and Joseph I. Goldstein Outstanding Junior Faculty Award from the ECE, and a UMass Distinguished Teaching Award Nomination.
Professor Yang earned his doctorate degree in the Materials Science Program at the University of Wisconsin at Madison in 2006, and his research specialty is nanoelectronics and nanoionics for computing and energy applications. His research has been funded by the Hewlett Packard Company, the Intelligence Advanced Research Projects Activity agency, the Air Force Research Lab, DARPA, and Goodix Inc.
According to the aims and scope expressed by Nature Electronics, this new Nature research journal is “interested in the best research from all areas of electronics, incorporating the work of scientists, engineers, and researchers in industry,” and their broad scope “ensures that work published reaches the widest possible audience.” (November 2017)