Carl Marziali
January 12, 2005 —
In the smoggy 1950s at Pasadena’s Jet Propulsion Laboratory, there were days when you could not see the mountain from the junior scientists’ trailers at the base of the Arroyo Seco.
But inside those trailers, a rare clarity of thought emerged creating an unlikely alliance of mathematicians and engineers. Bridging disciplines that used to mix as willingly as oil and water, a small progressive group laid the foundations of the digital age.
They were brilliant minds in their early twenties, fresh out of school and culled from every corner of the country. Later, most would join USC’s engineering school, turning it into a leading presence in digital communications and information theory. One would put his name on that school. At the time, however, academic honors were far from their minds. The race for space was on, and the Soviets were in first place.
William Lindsey, now a professor in the USC Viterbi School of Engineering, was an electrical engineering doctoral student at Purdue University in 1957 when the launch of Sputnik shocked America. Every 96 minutes the small aluminum can flew over the United States, and with every pass the country’s reaction grew more dire. What was next: nukes in space?
More intrigued that alarmed, Lindsey hooked up a radio to a speaker to see if he could pick up any signals. There it was: a steady beep in the ham frequency range, put there by the Soviets to make sure every American got the message. Lindsey knew the signal came from Sputnik, because it was Doppler-shifted: that is, the beep rose in pitch as the satellite approached and dropped down as it receded, every 96 minutes, exactly.
“When I hooked the speaker up to this audio signal and we could hear the Doppler shift, the whole university came running. It spread like wildfire through campus. The whole student body came to listen to this tone,” Lindsey remembers, still wide-eyed.
He realized two things: anything involving space was going to be a hot area for career-minded scientists; and, if you could pick up a signal from space, you could also modulate that signal and use it to carry information. That moment Lindsey decided to become a communications engineer.
“I knew there was something there,” he says.
Soon Lindsey joined the trailer group, formally known as Communications Section 331 in JPL’s Division 33. It was home to the theorists, and had been their home for some time. Lindsey was one of its youngest members.
There was Eberhardt Rechtin, now an emeritus professor at USC, but then the long-serving head of Division 33 and in his words “chief architect of the whole shebang,” meaning the systems for space communications, networks and devices.
There was mathematician Solomon Golomb, chief of Section 331 and father of the shift-register sequences used universally in message coding. Golomb now holds the Andrew and Erna Viterbi Chair in Communications at the USC Viterbi School. His contribution goes beyond mathematics, however. Both at JPL and later at USC, Golomb demonstrated a keen eye for talent and team-building.
“He had collected a very striking group of mathematicians and mathematically oriented engineers,” says Thomas Kailath, who spent a year at JPL and is now professor emeritus of engineering at Stanford University. “Section 331 was actually a powerhouse.”
Kailath credits Golomb for having the vision to erase some traditional boundaries.
“In those days mathematics and engineering were still regarded as somewhat separate entities,” Kailath says. “This was one of the groups that helped bridge that gap and showed the power of mathematical reasoning and mathematical modeling. That was the main contribution of the place.”
Besides Lindsey, Golomb recruited Lloyd Welch, who with Leonard Baum developed the Baum Welch Algorithm an important used in speech recognition and other areas. He is now professor emeritus at USC. And he brought on 22-year-old Andrew Viterbi, who would later become the co-founder of cell phone giant Qualcomm. Today, Viterbi holds the Presidential Chair in engineering at USC. He and his wife Erna are donors of an endowed chair and the most generous naming gift to any U.S. engineering school. Modern wireless technology is based in part on the Viterbi Algorithm for error decoding, the mathematics of which Viterbi says he learned from Golomb.
“I landed at JPL in a den of mathematicians,” Viterbi says. What a contrast from his previous employer, Raytheon, whose engineers Viterbi remembers as being “rather dismissive” of mathematical theory. Viterbi himself was right in the middle: a prolific thinker with an innate nose for the most fruitful intersections of theory and practice. Lindsey, who remembers himself as a naïve theoretician when he began working at JPL, credits Viterbi for redirecting him to problems that were not only intriguing, but important. And just in time: Viterbi arrived in June of 1957, three months before the launch of Sputnik turned the place upside down.
While Sputnik is credited with giving the U.S. space program a much-needed kick in the pants, few realize that JPL was ready to launch a satellite a full year before the Soviets. Just like Sputnik, the American version would have been small and rudimentary and, most importantly, first in space.
In September 1956, Golomb recalls, the directors of JPL and a sister group in Alabama brought the satellite proposal to the White House. It was discussed by scientists on President Eisenhower’s committee for basic research. That committee was supporting a rival project, called Vanguard. In addition, a majority on the committee felt that the exploration of space should be a non-military activity. Since JPL was run by the Army in those pre-NASA era, its proposal was dead on arrival.
“The Eisenhower administration was not aware that we were in any kind of a race at that point,” Golomb says. The launch of Sputnik raised their awareness level substantially. Within three months the government was ready with its counterpunch. In December of 1957, the U.S. invited the world’s press and television cameras to witness Vanguard’s debut at Cape Canaveral, Florida. Embarrassingly, Vanguard blew up on the launch pad.
“And that was our competition,” says Golomb.
A month later the scientists at JPL had their own answer to Sputnik, called Explorer I. Much to the Eisenhower administration’s relief, this one made it off the ground in one piece on January 31, 1958. There were clenched fists and cold sweats in the Pasadena control room when several tracking stations around the globe failed to detect the satellite at the expected times. The communications team later realized that the thrust of the Explorer I boosters had been greater than planned, making its orbit longer and therefore delaying its progress by about 15 minutes.
The ultimate barometer of national enthusiasm – Life magazine – featured photographs of Golomb and Viterbi in the control room. More importantly, the satellite’s success ensured the work of Section 331 would have far-reaching impact. By helping the satellite and mission control to communicate reliably, the scientists in the trailer had solved much more than the problem at hand. They had been drafting the blueprint for modern wireless communications.
The basic achievement had been the transport of information in a noisy environment with low-power transmitters. Out of that came satellite communications solutions, coding technology, the ability to track vehicles in space and time, the development of masers and rubidium and cesium standards for frequency and timing in all types of communications. It was a time of wide-open frontiers.
“What was our mission? We were defining it as much as we were being told what it was,” Golomb recalls.
The science behind the group’s achievements boiled down to two essential concepts. One was the idea of coding with shift-register sequences, developed by Golomb and further refined by himself, Viterbi and Welch. A shift register of n slots (n can be any whole number) is a device that takes an incoming stream of binary bits – ones and zeros – and, as each bit progresses from one slot to the next in the register, alters the message according to a predetermined formula. As each bit comes out of the register, it is fed back into the other side a certain number of times. The end result is a much bigger stream of bits with two advantages: it is coded to prevent jamming and, because it is redundant, with every original bit expressed multiple times, it is highly resistant to interference. Even if noise wipes out many bits in the stream, enough copies will be left to reconstruct the message. At the receiving end, a decoder inverts the coding formula and returns the original binary stream.
The other key concept was that of phase-lock loops, developed by Rechtin and Lindsey. The idea of a phase-lock loop is no different than every child’s favorite bath hobby: making the water slosh higher and higher until it splashes out of the tub. An oscillator on the receiver tunes itself to resonate at the same frequency as the incoming signal. Resonance is a natural amplifier, helping to pull the signal out of deep noise. “Phase-lock” refers to the need for the oscillator and the signal to be in phase and to stay that way.
“Phase-coherent communications and tracking is key to modern digital communications,” says Lindsey, who should know. While at JPL he worked secretly on the side for the U.S. intelligence community, monitoring Soviet communications. To his astonishment, he discovered that the other side was using phase-incoherent transmission – cheap, but very inefficient and not secure.
All these advances, though developed for communications in space, turned out to be vital for wireless transmission in a commercial airspace saturated with conflicting signals.
“I can say, without boasting at all, that we were the foundation of the global communications that we now have,” Rechtin says. Lincoln Laboratories, run by the Massachusetts Institute of Technology, also contributed important advances, as did Purdue and Stanford. The legendary Bell Laboratories was a bastion of research but, according to Golomb, focused heavily on the land-based telephone system.
For years, no one in the group grasped the significance of their research. “It was just interesting work,” says Welch. “I’m amazed, you see. We were basically in the digital communications era when nobody else was. Then as time goes on it’s found a use in commercial applications. They were using all the theory that we had developed back in the fifties and sixties.”
Years later, while consulting on the side, Lindsey was often asked how the group came together.
“I think God made this group,” he replied.
After God came Zohrab Kaprielian, who did his best to live up to his predecessor. Revered, scorned, loved, feared (he was all four at once), Kaprielian re-assembled the JPL group at USC. Nominally Kaprielian’s title was chair of the engineering department, but everyone knew he was much more.
Golomb recalls that at one point around 1970, there were five levels of administration between himself and the president of the university, “and all of them were Zohrab Kaprielian.”
“At that time we operated on the principle of one man, one vote,” Golomb says, “and Kaprielian was the one man who had the one vote.”
In matters of science Kaprielian used his vote brilliantly. Instead of trying to compete head to head with bigger universities, he focused on hot new fields where he knew USC could move faster than its stodgier rivals. Digital communications was one of those fields.
In 1963, on Viterbi’s advice, Kaprielian recruited Golomb from JPL. Golomb’s presence attracted others in the group: Welch and Lindsey, and later Rechtin and Viterbi (who had done his Ph.D. at USC because it was the only institution that would allow him to study while working at JPL). But Golomb’s eye for talent went beyond JPL. Also in 1963, he persuaded Kaprielian to hire Irving Reed, a gifted computer scientist then at RAND Corporation. Reed was best known for having built the first computer on the west coast, a desk-sized machine that humbled eastern rivals ten times its size.
What was more intriguing to Golomb was Reed’s research on error correction codes. Working with a collaborator, the late Gustave Solomon, Reed had shown that his algorithm for error correction was optimal, that is, unbeatable. At the time their finding was only of theoretical interest.
“Error correction coding was brand new,” Reed remembers.
Reed-Solomon codes became considerably less theoretical on the Voyager spacecraft. JPL launched Voyager in 1977 to explore the outer solar system. As historian Peter Westwick explains in his forthcoming book, Into the Black: A History of the Jet Propulsion Lab, 1976-2004 (Yale University Press, 2005), older error correction codes began to fail as the spacecraft sailed towards Uranus and Neptune. The Voyager project team then switched to Reed-Solomon codes, in combination with Viterbi decoding. The results were stunning: crystal clear photographs of the outer planets invaluable to scientists and inspirational to the public. Norm Haynes, the Voyager project manager, called this telecommunications success “the finest technological achievement of Voyager: being able to get images back from three billion miles away.”
Reed-Solomon/Viterbi decoding – and the underlying Golomb codes -- have been standard on every spacecraft since Voyager, including the recent Mars Rovers and Cassini. And for billions of people, Reed-Solomon codes are part of everyday life: they are inscribed into every single compact disc and DVD sold in the world. (A real-world tip here from Reed’s graduate student Gregory Dubney: when you clean your CDs, don’t wipe in a circle, as that will erase the Reed-Solomon codes over time and actually make the skipping worse. Clean the CDs by wiping towards the center.)
Together, these six pioneers – Golomb, Lindsey, Rechtin, Reed, Viterbi, and Welch – put USC on the map for information science and contributed to the USC Viterbi School’s ascent into the top tier of engineering schools. Golomb, Reed and Welch won the prestigious Shannon Award, named for Claude Shannon, the first formulator of information theory. All six belong to the country’s most select engineering society, the National Academy of Engineering. Golomb, Viterbi and Stanford’s Kailath are also members of the National Academy of Sciences.
Despite their many achievements over the last four decades, members of the old JPL group look back on those days with a special fondness.
“It was the finest adventure we ever had, that exceeded anything I’ve done since,” says Rechtin. “I don’t think we could have done it singly, none of us could.”
Says Lindsey: “So much work, so many things, so many areas started right in that Division 33, Section 331. We didn’t know where we were headed. A cluster of the key guys in digital communications came together and worked together closely as friends, as colleagues.
“We sorted out the problems and solved them and someone started using the results. Almost anything we solved we could write this paper and it would be published, because it was that new.”
That heady time may have come and gone in digital communications, but it repeats itself in emerging fields such as biotechnology, nanotechnology, bioinformatics and quantum information theory. Under Dean C. L. Max Nikias, the USC Viterbi School of Engineering has continued Kaprielian’s legacy by seeking out the brightest minds in the newest areas. In less than three years the number of tenure-track faculty has increased from 140 to 170 while the school as a whole now ranks sixth, tied with Caltech, according to U.S. News & World Report.
Many scientists over the ages have built new industries. The men of Section 331 can also say they helped to build an institution, the USC Viterbi School of Engineering.
—Carl Marziali
