As director of The Institute for System Level Integration (ISLI) based at Scotland's Alba campus in Livingston, Steve Beaumont sets curriculum and guides students at an academy that is being tailored to drive long-term system-on-chip success in Scotland.

Four of Scotland's leading universities-Edinburgh, Glasgow, Heriot-Watt and Strathclyde-formed the ISLI in support of a Scottish system-on-chip initiative launched with Cadence Design Systems Inc. (San Jose, Calif.) in the late 1990s.

As part of that initiative, Cadence set up a design services base here, now part of Tality Inc. At the same time, Scottish Enterprise, the national development agency, directed an initiative to convince Cadence to stay and for others to come to the region. As a result, the universities now feed graduates into ISLI, which is intended to help provide a flow of well-qualified engineers, as well as create a world-class environment for SoC research.

The Institute is only in its second full academic year but its gaining in confidence.

"What the industry is looking for now is the engineer with understanding from the top to the bottom," said Beaumont. "In the past it was possible to layer engineers with specialized knowledge, but now an architect or a systems engineer must have a general appreciation for the lower layers of abstraction," he said.

"So I suppose on that basis we need renaissance engineers rather than specialists."

According to Beaumont, the best graduates for the institute are those with a good grounding in both the electronic engineering and computer science disciplines rather than those who have stayed on one side or the other.

The universities do mix these courses, he said, but relatively few students take the option; the majority of applicants are schooled in one discipline or the other. The struggle, he said, seems to center on a culture that separates hardware- and software-oriented people at entry to university.

Steve Beaumont leads engineering doctorate and MSC students working on SoC at the Institute for System Level Integration in Scotland.

"We assumed we could take in students with hardware definition language (HDL) and programming skills. We want people who understand software overlaid on hardware and the interactions between them; people who understand issues like minimization of power consumption," Beaumont said.

"We've been disappointed at the levels of knowledge at the universities, he said. "A lot of graduates have not been introduced to the C language, for example. The C language or Java is pretty much an entry requirement for us." As a result Beaumont has found it necessary to send some students on crash courses in C.

ISLI started with eight master's students in 1999 and doubled its intake in 2000 with 16. Another 16 students have started a four-year engineering degree at the ISLI.

"For the masters course, we only have our students for a year," Beaumont said. "So there's a limit. We can't ram them with knowledge. We're not going to be able to turn a computer scientist or a software engineer into a very good analog engineer."

However, Beaumont pointed out that some industry trends tend to favor software engineers.

"CoWare/SystemC is a better route for a software engineer to find out about hardware," he said referring to one of the systems-level EDA companies and the C++ modeling standard being developed by the Open SystemC Initiative that it has helped to develop and promote.

"It's language based, and it's very familiar."

Beaumont contrasted SystemC with some other languages and graphical design environments such as Handel-C and MatLab.

"Ian Page might even say you don't need to understand hardware at all," he said, referring to the man responsible for developing Handel-C at Oxford University's Computing Laboratory in the early 1990s before encouraging its transfer to a spin-off company, now called Celoxica Ltd. (Abingdon, England).

However, as well as endowing aspiring electronics engineers with high-level language awareness, Beaumont reckons ISLI faces a challenge in building the computer scientists interest and confidence in hardware details. "Not all tools are going in that [higher abstraction] direction," he said. "And tools are not automatic."

The master's course includes modules on VLSI design, system partitioning, IP authoring, IP integration, bus architectures, verification, test, embedded software, real-time software.

While that represents a view of what system-on-chip engineers need in their portfolio today, ISLI also tries to prepare for the future, even though some of the technological twists and turns of that future are not yet clear.

"Perhaps we have to look at the communications issues of SoC, getting signals from A to B," Beaumont said. He explained that just as today's SoC designs integrate the architectures of yesterday's PCs and mobile phones, tomorrow's SoC designs will integrate today's systems; researchers need to think about how to put a portion of the Internet on a single-chip. "But that doesn't mean that the Institute should teach the detail of, for example, TCP/IP," he said.

On-the-job training

"The technology develops at such a fast pace, you can only learn detailed engineering on the job. You can't do it in a post graduate course. All we can do is introduce the students to some examples."

Beaumont gave highly parallel systems as a related example: "It's hard to teach because a lot of issues have not been resolved yet, but we can teach general principles. A microcontroller and a DSP core: at some level that's the same as 50-plus processors," he said.

But what are commercial companies interested in, such as those moving into Livingston as part of project Alba? "Verification, C++, Java. Some are interested in RF. But most of all companies are looking to us for people with system design skills."

Beaumont said that one area of research focuses on how to develop protocols that help put IP cores together. Just as industry uses protocols to link complex systems together successfully, so engineers could use the approach to link complex IP cores together on a chip. "There are topics like 'feature interaction analysis coming up in software' and 'how to protect the system from rogue IP' that could be useful," said Beaumont.

How does Beaumont see the SoC discipline in the commercial world?

"Industry does a good job, but I'm not sure they are as aware of the best technology as they should be. One of the weaknesses of industry is that it doesn't try out new ideas. Perhaps that's an area where we can help."

"We're about to start a substantial research project on bio-chip and system technology. We want to combine university expertise in interface technologies with SoCs." Beaumont said he also wants to make ISLI a hardware microelectromechanical systems (MEMS) center and contribute to system modeling and system design automation.

"The ultimate goal is to have a vibrant research activity at ISLI with leading-edge SoC technology. And bio-systems is a technology with a lot of potential," said Beaumont.

BACK TO THE FUTURE