4/28/2015 Ashish Valentine, ECE ILLINOIS
Written by Ashish Valentine, ECE ILLINOIS
As you load an article on your computer, its processor is running thousands of calculations in the background, crunching numbers for tasks like continually updating programs, running operating system processes, and maintaining firewalls.
If pushed too hard, a processor can overheat, leading to the machine failing. As computer engineers manage to squeeze more electronics on smaller chip sizes, these chips start accumulating more heat. Especially prone to overheating are the copper interconnects that run throughout the length of a chip, channeling the flow of electrical current.
In order to design processors, chip manufacturers need to know how much heat the processor can take, and what the conditions are like while the chip is running. Short of repeatedly building experimental chips, they need a cheaper way to find these out, so engineers turn to simulations.
This is where ECE graduate student Tianjian Lu comes in. Lu won the award for Best Paper at the Applied Computational Electromagnetics Society’s conference in Williamsburg, Va., for a paper that demonstrated an electrical-thermal simulation to virtually examine the effects of electricity and heat on chip interconnects.
“We import the structure of these devices into the simulation, and examine how electrical behaviors and the influence of changes in temperature can combine to affect chip performance,” Lu said. “Most of the state-of-the-art simulation tools are limited to understanding the design through a single physics in isolation, for example, they'll focus on just simulating heat physics or electricity physics. The design problem itself, however, features all these types of physics interacting together, so an accurate simulation has to account for that.”
Lu noted that the most challenging parts of his research were often implementing the real-world science he understood into code for his simulation. He noted that this part of the process required “calmness and patience,” as bugs in the code could take hours to find. Even if the code ran flawlessly, it could generate the wrong kinds of outputs, and Lu is all too familiar with the frustration of trying to find these invisible errors. However, Lu sees the frustration and continued effort as necessary to earn recognition.
“To that, I totally agree with what has been told by my advisor that if the work was smooth, then someone else would have already figured it out,” Lu said, chuckling.
Lu has submitted several two papers to ACES 2015 and this is his first award for Best Paper. Lu regards it as the culmination of years of hard work and a polishing of his skills through repeated effort.
“I am very grateful to my advisor who leads me into this field and guide me through all the difficulties encountered along this research," he said. "Without him, none of this would be possible.
“When my paper made it to the last top 10 competing for the award, we all had to give presentations on our work,” Lu said. “The most important thing is to be confident in your work, and I had coded every single bit of this on my own. It felt really good to be able expect questions to be asked, and to be able to explain every bit of this software to the group.”
Lu’s adviser, Professor Jianming Jin, was “very delighted” that his graduate student won the award for Best Paper, and was proud that a successful, energetic student of his was recognized for his efforts. Jin was especially impressed with Lu’s computational skills, especially with how Lu is able to not only find ways to improve upon high-frequency integrated circuit designs, but also implement computational methods to solve them.
“Tianjian has been a very constructive and productive member in our group,” Jin said. “He's always very active in seeking opportunities for innovative research with high potential for engineering applications. Given the novelty and quality this work and his excellent presentation, he clearly deserved this award.”