In 1946, a machine capable of performing complex mathematical equations could only be accommodated in a room the size of a gymnasium. Today, a chip with comparable computing power can fit on your fingertip.

The Electrical Numerical Integrator and Computer (ENIAC) may have launched the information age, but it didn't offer the convenience of laptop computing. Built at Penn during World War II, the machine took up 1,800 square feet and weighed a whopping 30 tons. Not exactly portable.

Until now.

Electrical-engineering students Lin Ping Ang, James Tau and Titi Alailima are designing a less-cumbersome version of the ENIAC. Together, they are creating a silicon chip that calculates like the world's first computer.

When the students began work last spring, they expected to complete the chip in time for this month's ENIAC festivities. "We thought, 'ENIAC is the world's very first computer,' " Mr. Tau remembered. " 'How hard could that be?' "

As it turned out, very. ENIAC was much more intricate than the students had ever imagined.

"I had originally thought that the ENIAC must be so simple that I could handle the task all by myself," Mr. Alailima admitted. "Ten months later, with a core team of three and several people doing supporting roles, the project has still been daunting."

"Even though the circuits themselves weren't very challenging, per se, putting its architecture on a silicon chip requires some of the most state-of-the-art, cutting-edge research on network connections," Mr. Tau added. "So it turned out to be a boon for us in the sense that we got to learn some of these issues and the things required to solve them."

The chip project was the brainchild of Jan Van der Spiegel, the undergraduate chair of electrical engineering, and Fred Ketterer, associate professor of electrical engineering. At the time of the idea's inception, Mr. Ang (SEAS'97) was doing research with Dr. Van der Spiegel. "When ... ENIAC on a chip came up, he approached me and said, 'Would you like to get involved in this?' " Mr. Ang recalled. "And I said, 'Sure, why not?' "

Dr. Van der Spiegel also asked Mr. Tau, a second-year graduate student, to chip in. "I took a class in VLSI [very large scale integration] design--a circuit-design class--and Dr. Van der Spiegel was looking for somebody to do the design work," he said.

"I had an interest in doing my thesis on VLSI design work," Mr. Tau continued. "Dr. Van der Spiegel gave me a topic my first year, but I hadn't done anything with it. Then when he had the chip idea, he asked me if I had done anything with the original topic, which I hadn't. He said, 'Great, here's another opportunity if you are more interested.' And sure enough, I was."

Mr. Alailima, another second-year graduate student, was interested in designing a chip version of the ENIAC before the project was even announced. "I actually had come up with the idea independently and talked to Dr. Ketterer about it, at which point he informed me that such a project was also being planned by various higher-ups," he recalled. "It was an exciting coincidence."

Before designing their chip, the students literally went back to the drawing board. They looked through archives and found the ENIAC's blueprints and manuscripts. They read through the material to gain an understanding of how the ENIAC worked, then they began to convert the machine's vacuum-tube circuits to analogous modern transistor circuits.

"At certain level of abstraction, there are similarities among the two types of circuits, although at the lowest level, similarities disappear," Mr. Tau said. "This is useful because once we identify the correspondence between a block in the ENIAC and modern digital component, we simply substitute the modern component for the ancient block."

"We just think about the functionality of it," Mr. Ang added. "Then, we try to keep as close as possible to the original architecture and transfer it to modern-day technology."

When the older components don't have a modern match, the students design their own circuits using CAD (computer-aided design) tools. The CAD tools are also used to test the circuits before the chip is manufactured.

"This step contains two parts," Mr. Tau said. "The first is essentially writing programs that describe the structure of the circuits, and running them like one would run a program on the PC. When we have verified that everything works, we then simulate the circuits at the transistor level, which takes into account the effects of such parameters as temperature, size of the transistor, capacitances and resistances ... . If this stage shows outputs that match the previous simulation, then we begin placing the transistors, using our computers, as they would be located on a silicon chip--a process called layout."

The students must place 300,000 transistors on the chip. And that's the easy part. The hard part is finding a way for the chip to mimic the ENIAC's 20 accumulators--the brains of the machine. "Primitive and rudimentary as the accumulator is compared to the modern microprocessor, it is, however, similar in function," Mr. Tau said.

Since the accumulator and microprocessor share similarities, the answer to the design dilemma may seem obvious: Build a chip with 20 microprocessors, one for each accumulator. But the solution isn't quite that simply. A normal chip isn't big enough to hold 20 microprocessors.

On the ENIAC, each accumulator communicated with the others. Because of its size, the chip will route information in a different fashion. Its "accumulators" will have restricted connections.

"This is similar to a telephone switching board, except on a silicon chip," Mr. Tau explained. "In the real world, the telephone company can build buildings as big as they want to encase their switchboards. They don't have any real area constraints. But on the silicon chip, that's a real big issue. So we have to figure out a way to minimize the connections, the switching, to achieve the level of communication we want between modules."

The chip may not be able to replicate the ENIAC precisely, but it will come close. "On the silicon chip, we are limited to the number of accumulators that one accumulator can communicate with," Mr. Tau said. "In that sense, the class of problems that our ENIAC can do may be limited, but we want our chip to do at least what the original ENIAC did, which is calculate the exterior ballistic equation."

The design of the chip will be finished in a couple of weeks, when it will be sent to a silicon foundry for fabrication. On Feb. 14, ENIAC's 50th anniversary, the students will be able to show the chip layout and the design of the different building blocks that constitute the full ENIAC.

The silicon foundry will need a few weeks to manufacture the chip. In the meantime, Deborah Seider, another electrical-engineering student, will keep busy writing software to program it. The software interface is her senior design project.

"We intend to build an ENIAC kit that can be taken to museums and high schools to show how ENIAC works," Mr. Ang said.

Mr. Ang, along with Mr. Tau and Mr. Alailima, admits that they initially underestimated the sophistication of ENIAC. Still, they're learning a great deal from their experience, and they've gained a new appreciation for the scientists who built the world's first computer.

"One of the reasons why our goal was so ambitious at the start was that we didn't know very much about the ENIAC," Mr. Tau said. "In other words, we learned as we went along. ... So, in a sense, this is sort of like a research project, not just a recreation of something a bunch of brilliant people did 50 years ago."

"I've developed a lot of respect for the ENIAC," Mr. Alailima added. "It's a lot more than I initially thought it was. Modern hand calculators may have it beat for speed, but the ENIAC incorporated lots of features that disappeared from computing for a long while, but are now reemerging in cutting-edge technologies."

Return to Compass Features for February 6, 1996