Henry Samueli: Champion of Digital Broadband

Before the Cable Modem—Way Before

Samueli started down the path that would lead to cable modems when he was in middle school. But he wasn’t thinking about a future career when he enrolled in an electric shop class. It was just that, he says, “electricity seemed kind of mysterious, compared with metal or wood.”

The teacher assigned a crystal radio project, he recalls, “but wrapping a wire around a tube from toilet paper and connecting the wire to a crystal wasn’t that exciting to me.” So he thumbed through an electronics catalog looking for an alternative. A
Graymark five-tube radio caught his eye. It took some convincing before the teacher agreed to let him tackle the project, which came with complicated instructions and involved learning how to solder.

“I worked every night,” Samueli says. “There were hundreds of connections that I had to solder up. It took a full semester to build it, and, at the end, I brought it into class, plugged it in, and sound came out of it. I was totally blown away. And I literally made it my mission in life to figure out how radios work.”

Samueli’s teacher was blown away as well. And what he said crystallized Samueli’s future. “He told me, ‘Henry, honestly, I never ever thought you could do this. But clearly, you’ve got some special gifts. I think you should pursue electrical engineering as a career. You’re going to do something big someday.’ ”

UCLA Takes Hold—and Never Lets Go

Samueli eventually applied to UCLA—a university with a good electrical engineering program and affordable tuition that was close to home. He went straight through to a Ph.D. but, he says, didn’t really understand how radios worked until a few years beyond that.

After collecting his Ph.D. in 1980, Samueli joined TRW to work on defense communications projects. He says he loved every minute. “It’s a tremendous opportunity to learn because you’re dealing with superhigh tech, the greatest technology at the time. And with a big budget, you can build very sophisticated things,” he says.

Samueli didn’t completely leave the world of higher education. In his spare time, he taught a circuit-design class at
California State University, Northridge, and then several circuits and signal processing classes at UCLA. In 1985, UCLA offered him an assistant professorship, and he left TRW, taking coworker Henry Nicholas with him as his first Ph.D. student.

The two formed the core of what would become the multidisciplinary communications research program in UCLA’s Integrated Circuits and Systems Laboratory. They collaborated with several faculty members in the electrical engineering and computer science department to develop digital modem chips.

Broadcom cofounders Henry Samueli [left] and Henry Nicholas pose in front of the company’s headquarters in Irvine, Calif., in 1999. Ted Soqui/Corbis/Getty Images

“Chip design is a very complex and broad discipline,” Samueli points out. “There are analog designs, digital designs, multiple systems, various architectures. While such a multidisciplinary approach is standard today, it was fairly unusual at the time.”

AT&T Bell Labs was leading the world in digital-communications research, Samueli recalls, using low-speed modems that communicated in the same bandwidths as the human voice. The labs built those modems using programmable digital signal processing chips from
Texas Instruments and others.

“It was a software-driven approach to building digital signal processing,” Samueli says. “And it only ran at data rates of tens of kilobits per second. Our challenge was how to take those algorithms and make them run at tens of
megabits per second—one thousand times faster.”

Samueli and his colleagues concluded that a programmable architecture using software was just too slow. So they began investigating parallel architectures that could implement sophisticated algorithms on a single chip with no software, just dedicated hardware. “That was our innovation. Back then, it was very novel,” he notes. “Today, it’s what makes AI processors work.”

UCLA researchers who specialized in analog signal processing collaborated with the group to integrate high-speed analog-to-digital and digital-to-analog converters into the core functions of the chip—“really breakthrough work,” Samueli says.

“I was totally blown away. And I literally made it my mission in life to figure out how radios work.”—Henry Samueli

Samueli and his team weren’t thinking patents while they were doing this research. As academics, their focus was on publishing their results—some 100-plus papers over 10 years. But many others saw commercial potential in their work.

“After we’d publish a paper, we’d go to a conference and make a presentation,” Samueli says. “People would come up to us after the talk and say, ‘This is really neat stuff. Have you ever thought about commercializing it and starting a company?’”

Samueli and Nicholas took the leap in August of 1991, incorporating Broadcom Corp. and chipping in US $5,000 each to rent an office and buy computers and office supplies.

Samueli kept working full-time at UCLA while Broadcom began bringing in small defense contracts: developing a digital frequency synthesizer for TRW, a digital filter for a Rockwell microwave radio, and, for the U.S. Air Force, a digital filter to protect GPS signals from jamming.

“These projects funded our R&D, and we gained more and more knowledge,” says Samueli. [For more on Samueli’s early career, see
this 1999 profile.]

Scientific Atlanta Connects with Broadcom

In December 1992, a student of Samueli’s
gave a presentation at Globecom (the IEEE Global Telecommunications Conference, that is) about a prototype 10-plus megabit-per-second digital modem chip the group had developed.

“What was different in their chip is that it integrated digital and analog,” recalls Leo Montreuil, then an engineer at
Scientific Atlanta and now an IC design engineer at Broadcom. At the time, Scientific Atlanta shared the U.S. cable TV set-top box market with only one competitor, General Instrument. “We had many companies making chips for Scientific Atlanta, but not that kind of chip.”

After the presentation, Montreuil approached the student, who referred him to Samueli. Montreuil met with Samueli and Nicholas three months later.

Henry Samueli is this year’s recipient of the IEEE Medal of Honor for his contributions to digital broadband technology and his support of STEM education.Peter Adams

Scientific Atlanta wasn’t just casually curious about the work. The company had signed a major contract with Time Warner to build 4,000 set-top boxes for the world’s first digital cable system, called the
Full Service Network. It needed a digital modem for that box, but the necessary chips weren’t commercially available.

“What they were trying to do in a single chip seemed so much better than multichip systems being developed by others,” says Montreuil. “When you go from analog to a digital implementation, you have to worry about drift, temperature sensitivity, and other issues. The more you can implement in the digital domain, the more predictable is the system.”

Scientific Atlanta awarded a $1 million development contract to Broadcom in June of 1993. Although Broadcom’s design ended up using three chips, the company did combine analog and digital circuitry on the same silicon.

“The project was straightforward,” Samueli says, “because it was based on the prototype designs we had already done. And it worked the first time, flawlessly.”

Time Warner’s digital cable network—activated in Orlando, Fla., in early 1995—was a technical success, but Time Warner didn’t take it any further. The network wasn’t intended to be financially viable, Montreuil says, pointing out that the core of each home system was a prohibitively expensive Sun SPARC workstation. “The goal was to acquire knowledge and to get our foot in the door for the next generation.”

Broadcom’s modem design impressed both Scientific Atlanta and General Instrument. The two competitors invested $1 million each, for a 10 percent total stake in the startup. That funding allowed Broadcom to keep working on digital modems, to reduce the cost by putting all the functions on a single chip.

Sherman Chen was a senior engineer at General Instrument at the time. “We knew then that the Broadcom device would dramatically extend the boundaries of communications,” recalls Chen, who is now vice president of engineering in Broadcom’s broadband video group. “Ideas like advanced error correction and digital compression were around, but they were all just elegant theories until Broadcom built the first mixed-signal silicon for broadband communications. Broadcom created an industry.”

Broadcom wasn’t the only company chasing the low-cost digital modem grail. One key competitor was LANcity, which had a $500 digital modem. The market was evolving quickly, and it was becoming clear to cable operators that this new technology would require standardization. Broadcom, CableLabs, General Instrument, LANcity, 3Com, and others
began collaborating in 1995 to create an international standard called the Data-Over-Cable Service Interface Specification (DOCSIS).

“People would come up to us after the talk and say, ‘This is really neat stuff. Have you ever thought about commercializing it and starting a company?’ ”—Henry Samueli