Researchers at universities across the country, including UCSB professors Kwang-Ting Tim Cheng and Malgorzata Marek-Sadoska, are developing tools to make designing and testing integrated circuits easier.
An integrated circuit is a series of semiconductors on a chip, each of which contains instructions for the computer in which it is placed. Having more complex circuits with larger numbers of transistors makes it possible to design faster computers, capable of carrying out more instructions at once.
Launched two years ago, the nationwide research effort achieved its original goal of creating a complex integrated circuit. It has now turned its attention to a bigger problem. The largest obstacle to creating integrated circuits in the semiconductor field is design productivity, Cheng said.
“There’s probably more than a hundred million transistors in a chip the size of a fingernail. Manufacturing technology allows us to integrate that many transistors within a single chip,” he said. “Manufacturing is still challenging, but the challenge is not as big as the design itself.”
The design of an integrated circuit needs to be completed before manufacturing can begin – a task that sometimes takes over 800 people to complete.
“The design itself needs to be done very quickly because you are waiting to take your plan to the market,” Cheng said. “Likely, the chip will take 12 to 18 months just to design. This translates to loss of money and loss of market.”
With the development of new technology and design methods, this phase can be cut to three months and require far fewer designers, Cheng said.
The new methods could fundamentally change how an integrated circuit is designed. The main idea is to automate a large portion of the process by creating an interface that will construct a design with less input from the designer than previously required.
“The designers won’t have to deal with millions of transistors,” Cheng said. “Instead, they use a small number of these high-level objects and have the tools and methodologies automatically translated to a large number of transistors.”
The UCSB research center is also focusing on testing chips.
“The manufacturing process is so complicated and it is not perfect,” Cheng said. “Determining whether [a chip] is good or bad is complicated. Just like driving a car; you test drive it for 20 minutes, it’s working fine. That doesn’t mean it will work well for years. The manufacturers have to make sure it’s working fine because they offer a three-year warranty.”
By automating portions of the design process, researchers also developed a way to make testing easier.
“We design the chip so later on it’s much more readable – so we know when the chip is fabricated if it’s good or bad,” Cheng said. “This process will reduce the cost of testing and also enhance the quality and reliability of the integrated circuit.”
Because circuitry research takes place at universities across the country, communication takes time, Cheng said.
“It’s very unique, because very few centers in the nation have this kind of setup,” he said. “In the past we’ve kind of been competing with each other.”
The project’s funding has more than doubled, to $8.5 million, since its inception. Approximately 75 percent of funding comes from industry and 25 percent from the military.