A Drexel professor was awarded a National Science Foundation grant May 18 for his development of wireless networks on microchips.
Baris Taskin, an associate professor in Drexel’s College of Engineering, has been researching how to implement wireless interconnects on microchips for several years. Kapil Dandekar, an associate professor and associate dean of research in the College of Engineering, is a main collaborator in the research.
Wireless interconnects are used with more than one wireless communication device, such as antennas or couplers, on the same integrated circuit system. This technology has the advantage of freely placing communication nodes on the chip as long as they are within the wireless communication range.
The wireless nodes replace the resource-demanding, wire-based interconnects. Wire-based interconnects are similar to the human intestine, where wired interconnections can have great length despite being condensed into a small space. This means to make a chip functional it must have a large volume of wired connections, which take up a large area on the chip.
“Unlike wire-based interconnects, wireless interconnects have broadcasting capability to communicate with more than one node at once, for instance to be used for a system-wide reset,” Taskin said.
Wireless interconnects allow communication over integrated circuits that would normally not be possible with wire-based connections due to the circuits’ layouts.
“This is the roll-call for wireless technology to establish wireless interconnects, which can be used instead of having bulky, long wires amongst the millions of other components on the chip,” Taskin said.
A hybrid version of the chip is a happy medium because both wired and wireless connections have their pros and cons. Traditional wire-based interconnects perform better over short distances because they do not have the power overhead or the footprint of a wireless communication node. There is also a point where wireless connections become favorable because devices that use more wire for wire-based connections cost more.
Wireless connections have the ability to perform better over long distances as long as nothing comes between communication nodes, which is the case in the crammed integrated circuit world.
“The hybrid use of the two is ideal given that an integrated circuit has millions of paths spanning a wide range of interconnections that need to be established,” Taskin said.
The communication bandwidth is increased a great deal compared to wires. Using wires when connecting multiple cores that are not close in proximity requires more area and resources, and the data would transfer substantially slower.
Bus technology, a collection of wires, is what is currently used to connect distant cores.
“But bus technology is not scaling to 16 or 32 core processors. We are working with 1,000 core systems, where the connections resemble a small network, a term coined ‘network-on-chip.’ Wireless interconnects are working wonders in improving communication over distant nodes of the network, while wired buses still rule the local communication between nodes in vicinity of each other,” Taskin said.
Two microchips were created for proof-of-concept studies, including a scaled-up version for fast prototyping. Scaling up is not a good thing in microelectronics, however. In order for a chip to house more than one wireless link, it would have to be between 10 and 20 micrometers on one side alone.
“These chips are not economically feasible on a research budget. Instead, we are perfecting the design of a single wireless link on the tiny chip and simulating scalability to the larger systems,” Taskin said.
The price of creating a simple four-square-millimeter microchip containing one wireless connection is almost $90,000. This price includes discounts given to educational institutions.
In Drexel’s very large-scale integration laboratory, the microchips are developed through a process called “tape-out.” The tape-out experience is seeing an integrated circuit design go from its computer-aided design form to silicon.
Taskin was awarded $400,000 to use to further the wireless microchip’s research over the next three years.
“Our agenda is to build two other chips within the next three years. While we lose some of these researchers to Silicon Valley, we bring new people on board,” Taskin said.
One new student has already been recruited to work on the research, and Taskin plans to hire more.