Kershner’s Roulette

You have to give yourself a chance to get lucky.

There’s plenty of time, if we work hard enough.

—Richard Kershner, Transit project leader

I

f people leave a legacy in the eyes and voices of those who knew them well, then Richard Kershner should rest easy Kershner led the Tran­sit team with an authority bestowed willingly by those who worked for him.

He instituted a policy of “cradle to grave” engineering, which means that the person who designs a particular component also has responsibility for testing it. This was not (and is not) a common practice, but Transit team members perceive it as having been critical to the project’s success. Kershner eschewed line management as much as possible. If he had a ques­tion about a particular component, he would go directly to the person building that component rather than the department head. The policy was applied widely. Someone working for Guier, for example, could and would go around Guier if he thought an issue would be discussed more profitably with someone else.

There were the inevitable personality conflicts and disagreements. Yet a widely held memory among the Transit team is that the optimum technical solution determined the outcome of an argument, not an indi­vidual’s position in the hierarchy. All recall that Kershner was the final arbiter. He would tell them, “you together have n votes, and I have n plus one votes.” Yet his decisions seem to have been accepted because of his character, not because of the authority vested in his position. His most admired characteristic was his acceptance of responsibility when things went wrong, and over the years many things did go wrong in the Transit program. Kershner would say that whoever did not like what had hap­pened would have to deal with him first. More than thirty-five years after the project started and fourteen years after Richard Kershner’s death, the Transit team’s respect is still palpable.

One can look for the reasons among articles Kershner wrote about management practice. They are as dry and irrelevant as, I suspect, most for­mulas for successful management. Perhaps they could not be applied by
someone other than himself; he seems to have had a natural tact and respect for his staff that won him great loyalty.

Immediately after McClure had explained his idea for satellite navi­gation, Kershner was brought in to help put together a proposal. He was joined by Guier, Weiffenbach, and Bob Newton, a reticent, elegant An­glophile. During Transit’s development, Newton was to contribute to nearly all areas of the physics. In future years, these four would commemo­rate their early collaboration by celebrating the anniversary of the first suc­cessful Transit launch.

Their initial proposal was completed by April 4, 1958. It was marked “confidential,” a low level of classification. Submarines were mentioned only in passing, even though Polaris provided the impetus for the proposal. Informal discussions about navigation for Polaris must have been going on at high levels of APL and the Pentagon.

During the next few years, details of the proposal would change. The ionosphere, for example, deemed at this time to be the biggest problem, was soon supplanted in importance by the need to understand the earth’s gravitational field. Nevertheless, a surprising amount of the original pro­posal was to remain.

For the first draft, Guier tackled computing, Newton dealt with the satellite’s motion, and Weiffenbach took on the ionosphere and assessment of error—frequency drift, for example. Kershner worked on the overall system design. After the first draft, they all worked on all sections, cri­tiquing and revising, guided by McClure’s intimate knowledge of what Special Projects wanted and would accept.

A worldwide “satellite Doppler navigation system” for all weather was possible, argued the proposal, because of the recent developments showing that it was technically feasible to establish artificial satellites in predetermined orbits with lifetimes measured in years. A position fix with a CEP (circular error probable—a circle of radius within which there is a 50% probability of finding an object) of half a nautical mile should be pos­sible immediately. CEPs of one-tenth of a nautical mile, about six hundred feet, were likely in the near future, and such accuracies, stated the proposal, were greater than those of any known military requirements. Transit team members say that the first operational satellite launched at the end of 1963 achieved a CEP of one-tenth of a mile. Even though there were still prob­lems to be solved, this was an impressive feat.

An alternative approach to navigation, involving further develop­ment of a coastal radio ranging system called Loran, would also have been accurate, but only when the submarines were within range of its signal. Loran, says a former British liaison officer to Special Projects, would have been too limiting for Polaris.

A particular strength, the proposal pointed out, was the passive nature of the proposed navigation system, which would not betray a submarine’s position. The submarine would need only to approach the surface and to deploy its antenna at night for about eight minutes.

Various approaches to calculating latitude and longitude were out­lined. One involved our old friend “miss distance,” that is, the range at closest approach. Given the miss distance, one could plot a line on a map representing positions for which the submarine might be at the corre­sponding range from the satellite. If this process were repeated for signals from two satellites, the latitude and longitude would be the place where the lines of position intersected. Conceptually, this was an approach famil­iar to navigators.

APL, of course, was not trying to persuade anyone that relying on miss distance was a good approach, and the proposal went on to describe in general terms the computational and statistical techniques developed by Guier and Weiffenbach during their research into orbital determination and to say that these would be applied for position fixing. The greater accuracy attainable by comparing the curves of the theoretically generated and received Doppler signals at many points was pointed out. They needed to make this argument explicitly because many opponents at the time, recalls Guier, though that the method was “out in left field.”

To recap, the Transit proposal was essentially this: A satellite in a known position would broadcast two continuous waves. A comparison of the two signals would allow the submarine’s computer to eliminate the effects of ionospheric refraction. That same computer would then generate—for dif­ferent latitudes and longitudes—the Doppler curves that the submarine would receive from a satellite in that position. The inertial navigation unit would provide an initial estimate of position to the computer, which would run through a number of iterations until it found the latitude and longitude that generated the Doppler curve that best fitted that received from the satellite. These values of latitude and longitude would then be fed to the inertial navigation system to automatically correct its output.

The system, which was yet to be named Transit, was explained through the example of a satellite in a polar orbit at an altitude of four hundred miles. Such a satellite takes ninety-six minutes to complete one orbit. During these ninety-six minutes, the earth turns through about twenty-four degrees of longitude. The point immediately below the satellite—the sub­satellite point—moves westward by an equivalent number of miles, depending on the latitude. Thus, how frequently one could “see” such a satellite during the course of a day depends on latitude. Near the poles, the satellite would be visible on every orbit; near the equator, less often.

Simply seeing the satellite, as Guier and Weiffenbach had come to understand by observing Explorer I, would not be enough. The relation­ship of the orbit to receiving equipment on the ground had to fall within certain limits. If the arc of the orbit were too distant, the recorded Doppler shift would change far more gradually with time than if the satellite were nearer. The quantity of positional information would then be less per unit time, and with constant noise being received, the so-called signal-to-noise ratio would be too low for optimum accuracy. Moreover, the signal would be passing through a greater distance and would be subject to more atmo­spheric refraction, which would obscure the more gradual Doppler changes of the distant satellite. On the other hand, if the satellite were to pass nearly overhead, the computations would result in an error in longi­tude. In between, there was a wide region in which the total navigational error would be acceptably small and would be insensitive to the geometry of the satellite pass.

These limitations were pointed out, but the proposal did not recom­mend a particular number of satellites or an orbital configuration. Never­theless, one begins to see how the laws of physics and the requirements of the job to be done, for example worldwide navigation versus position fix­ing in the arctic, combine to impose limits on the configuration of orbital inclinations and altitudes.

A vital part of the system would be the oscillators producing the continuous waves for the Doppler measurements and the reference frequency in the submarines receiver. The stability of the oscillators and the error in the frequency measurement defined the system’s theoretical limit of accu­racy—the limit achievable if all other problems were solved.

The first proposal asserted that the state of the art for oscillators meant that a CEP of a thirtieth of a nautical mile (roughly two hundred feet) was theoretically possible. Although the proposal had made clear that this accuracy was dependent on a number of assumptions, including a favorable geometry between satellite and receiving station, critics thought that APL was saying that it could achieve this number regularly in practice. When the figure leaked out, McClure had to field outraged and disbeliev­ing phone calls, including one from Bill Markowitz, head of the Naval Observatory. Eventually, Transit surpassed even this accuracy, but in 1958 the proposal’s claims seemed implausible.

The situation was not helped by the fact that many of the critics either did not understand or did not want to understand the computa­tional and statistical approach APL was proposing. Guier and Weiffenbach recall one group that converted their data to miss distances, calculated, inevitably, a very poor orbit, and so dismissed their work.

Transit’s history was similar to that of many technological develop­ments. First, Guier and Weiffenbach explored something new and unex­pected, initially with no clear idea in mind. Then they found a purpose for their research—orbital determination. In March, McClure entered the stage with a way of exploiting their research for a practical application. McClure brought with him Richard Kershner, someone who today would be called a systems engineer. Kershner supervised the development of a proposal, which, because it served a military need, won funding. It encountered opposition and was not widely accepted for some time.

One could write of this last aspect of Transit’s development, that there is “nothing new under the sun,” because the satellite’s development has much in common with the pursuit of longitude, the critical navigational problem that occupied physicists of the seventeenth and eighteenth cen­turies—some of these same physicists who laid the scientific foundations for Transit.

A diversion into history is too hard to resist. Geopolitical tensions between two superpowers—Spain and Portugal—provided the initial impetus for the pursuit of longitude. The two countries had a territorial dispute that could not be resolved because they did not know the location of the meridian in the Atlantic ocean that separated their claims to sover­eignty. Before a solution was found, there were political hearings, testi­mony from distinguished scientists, competing technical solutions, pleas for public funding, and suspicion of the new technology from naval officers. Once a solution to the longitude problem was found, it became invaluable to commercial shipping interests.

In the seventeenth century, the problem of determining longitude came down to the need for a means of telling the time accurately at sea. If a clock of some kind were set according to time at a reference longitude, say Greenwich, and local noon occurred at 2 RM. Greenwich time, then you knew you were thirty degrees west of Greenwich. (The earth turns through fifteen degrees of longitude per hour).

Telling the time at sea was acknowledged to be very difficult. On land there was no problem. Pendulum clocks were accurate to within a few seconds, but they did not work well on the pitching and rolling deck of a ship. First the Portuguese, then the Spanish, the Dutch, and the French offered rewards for ideas leading to a method to tell the time at sea. Every crank in Europe responded. Overwhelmed by dubious proposals, Spanish officials returned a promising idea from an Italian named Galileo Galilei that suggested taking advantage of the times of eclipse and emergence of the moons of Jupiter. Given the difficulties of stabilizing a pendulum at sea, the idea was good. But it languished in bureaucracy.

A century after the Portuguese first tackled the problem, the English Parliament bestirred itself in an effort to placate a public scandalized when, in 1707, Admiral Sir Clowdisley Shovell led the fleet onto rocks in bad weather off the tip of Cornwall. Some two thousand men died. Every nav­igator but one had agreed that the fleet was southeast of its actual position just off the coast of Brittany.

After more losses of men and goods, a petition signed by several cap­tains of Her Majesty’s ships, merchants of London and commanders of mer­chantmen was set to the House of Commons in 1714. Though couched in florid, formal language, its message was clear: Parliament should act. The petitioners suggested that public money be offered as an incentive to any­one able to invent a method of determining longitude at sea.

Parliament sought expert advice, calling, among others, Isaac New­ton and Edmund Halley to testify before the parliamentary committee of the House of Commons on June 11, 1714. Newton was not encouraging. He spoke of “ . .. several projects, true in practice, but difficult to execute.”

Rather than celestial schemes, he favored the development of an accurate clock, telling members that of the various options for determining longi­tude, “One is by watch to keep time exactly, but by reason of the motion of the ship at sea, the variations of heat and cold, wet and dry and the dif­ference in gravity in different latitudes, such a watch has not yet been made.” He offered no engineering advice.

A few weeks after this testimony, Parliament enacted legislation offering a reward to the first inventor to develop a clock that met specifi­cations laid out in the act. The award was £20,000, which for the time was a considerable amount of money. The act also established a Board of Lon­gitude, which was given the task of awarding smaller grants for promising ideas as well as responsibility for determining when the terms of the act had been met. An unschooled but skilled cabinetmaker called John Harri­son received several of these grants and eventually developed a chronome­ter (known among the cognoscenti as “chronometer No 4”) that fulfilled the act’s specifications.

The board proved reluctant to pay the full £20,000. There were conflicts of interest, ownership disputes, and numerous sea trials. Unde­terred, the Royal Society honored Harrison, but the monetary reward was still not forthcoming. Then the board paid up in part. Unsatisfied, Harri­son appealed to the king, who offered to appear under a lesser title to argue the case before the Commons. Eventually, the board paid out the full amount.

Another time, another place: In the late spring of 1958, Kershner prepared to do battle for what would be a controversial new approach to navigation. Gibson, McClure, and Milton Eisenhower, the president of Johns Hopkins University and brother to the other president, were playing their parts too.

Kershner made several trips to the Pentagon, often taking Guier and Weiffenbach with him. At first, Guier and Weiffenbach felt apprehensive and out of their depth. Amidst these efforts, Weiffenbach was awarded his Ph. D., something of a sidebar as APL fought competing navigation proposals and criticism. The Jet Propulsion Laboratory, which had a deservedly high repu­tation, posed a particular threat when it challenged APL’s error analysis. Eventually, there came a meeting at which Guier and Weiffenbach argued for an opportunity to fly an oscillator on someone else’s satellite. This would give them a chance to prove the concept and to show that an ionospheric correction was possible. Their audience, though, had already been convinced of the validity of APL’s approach to developing a navigation satellite, and Guier and Weiffenbach left with the promise of funding for a satellite, not just an oscillator. (An oscillator was flown on the Department of Defense’s Discover satellites as part of the effort to determine the earth’s gravitational field). This happened sometime in the summer of 1958, and APL had by now developed technical plans for eight experimental satellites. Their aim was to have a prototype operational satellite aloft by the end of 1962.

Others had joined the project since April. In July, Weiffenbach sent John Hamblen the operating characteristics for the satellite’s antenna. Weif­fenbach was also designing the first oscillator and reviewing what was known about the ionosphere. Harold Black had his wish and was working for Guier and Newton on the computational effort for orbital determina­tion and prediction. Joy Hook, a skilled programmer, had joined the team and was rapidly becoming invaluable. Many subroutines were being com­pleted, including those for organizing the recorded data on magnetic tape, for applying the refraction correction, and for curve fitting. There was, says Black, a quiet desperation about much of the software effort. Very few people understood it well.

In November 1958, three of APL’s engineers visited Iowa State Uni­versity, seeking advice from George Ludwig in James Van Allen’s group about the kind of environment their satellite would encounter in space. They returned with information about how ISU built and tested its satel­lites and about temperature control. They picked up schematics for the Explorer III satellite, which had been launched on March 26, as well as information about the satellite’s radiation observations. These observa­tions, which were the first data supporting the hypothesized existence of radiation belts, had not yet been widely circulated. They were to be important to Transit because the satellites would be in orbits that took them through the newly discovered radiation belts.

As work on Transit progressed, more detailed questions emerged. How many transistors, for example, would be needed in the clock circuit or the oscillator circuit? The original proposal of April 4 had said that the satellite would be fully transistorized, so these questions were significant, especially at a time when a transistor could cost as much as ninety dollars. How much power would the circuits need? How would they be packed and into what volume? The original specifications were for 293 transistors

Researchers from the Johns Hopkins University Applied Physics Laboratory position a satellite from Iowa State University on top of the Laboratory’s Transit 4A satellite in June 1961. Courtesy of Applied Physics Laboratory.

or diodes. With an estimated packing volume of two cubic inches per tran­sistor or diode, the electronics could not be packed into a volume of less than 586 cubic inches and would weigh eighteen pounds. During the rest of 1958 and through 1959, these numbers were refined.

Another difficulty arose—how to affix the electronics to the circuit boards. After some faulty starts with soldering and some burned-out tran­sistors, the engineers decided to weld the components. But different coef­ficients of expansion between the connections and the components led to many breakdowns. Reliability did not really improve until the advent of integrated circuits in the mid 1960s.

On December 15, in the midst of these ongoing problems, the Advanced Research Projects Agency (ARPA) awarded APL full funding of $1,023,000 for the Transit project. Although ARPA awarded the contract, the work was sponsored jointly by the Special Projects Office and was administered by the Navy’s Bureau of Ordnance.

The aim was for Transit to be operational by the time the Polaris submarines were on station. This goal was not met. Two submarines were on patrol by the end of 1960, three full years before Transit was opera­tional. But the Polaris deployment had moved ahead of schedule, and the Transit development slipped by a year. Even so, Transit’s development went ahead at a breakneck pace.

The objectives initially fell under two headings: the Transit program and NAV 1—the satellite. The Transit program was defined as the devel­opment of a shipboard navigation system, work on the earth’s gravitational field, and a feasibility study of a Doppler navigation system for aircraft.

The first aim is obvious. The second became critical. The last aim, however, was curious for two reasons. First, single-satellite Doppler naviga­tion was never going to work for aircraft because they need to determine altitude as well as latitude and longitude and thus would need to record either two passes of the same satellite or one pass each of two different satellites. For aircraft moving at, say, eight miles a minute, this was not an option, and the higher echelons of the Pentagon had decided to develop inertial navigation for aircraft. However, there was a new requirement for all-weather navigation for the aircraft of the radar picket fence, which stayed airborne for eighteen hours at a stretch. The thinking was that Transit coupled with interpolation from inertial navigation units might serve these aircraft.

The second part of the contract—Nav-I—comprised three satellites, which were to be delivered to Cape Canaveral in time for a launch sched­uled for August 24, 1959. Once the first satellite was in orbit, its nomen­clature would change from Nav-1 to Transit IA. NAV-Vs objectives were to demonstrate the payload (another word for satellite instrumentation), the tracking stations and data processing, and the Doppler tracking meth­ods, as well as to collect data (from analysis of Doppler curves) about the shape of the earth and its gravitational field.

The plan at the end of the 1958 was that the NAV-1 satellites would weight 270 pounds and be launched on a Thor-Able rocket toward the end of August 1959. This was considerably heavier than the original pro­posal’s estimate of 50 pounds, but Transit IA was an experimental satellite. Kershner later insisted, in the face of considerable opposition from members of the Transit team, that the weight of the operational satellites be lessened to make them compatible with the less costly Scout rockets. Some of the Transit documents suggest that the aim had always been to make the oper-

An early tracking station for Transit satellites. The station located at the Johns Hopkins University Applied Physics Laboratory was for research. Courtesy of Applied Physics Laboratory.

ational satellites compatible with Scout. Team members remember that they wanted to stay with the heavier-lift launch vehicle. As was often the case, Kershner won the debate. But the transition proved more difficult than anticipated and held up the program for a year, delaying the launch of an operational navigation satellite until the end of 1963.

By February 1959, APL had made good progress with the satellite antenna. The following month, a year after McClure had called Guier and Weiffenbach into his office, seventy percent of the satellite’s electronics had been built and mechanical fabrication had begun. Results of analytical work suggested that by 1963 a simplified Doppler navigation with an accuracy of one mile could be performed by hand, and that with more sophisticated techniques (curve fitting), an accuracy of one-tenth of a mile would be possible.

In the spring of 1959, the first computer program (ODP-1) for determining satellite position was nearly complete. It took the Univac ПОЗА twenty-four-hours to run the program in order to predict an orbit for twenty-four-hours. A year later, when the lab installed the IBM 7090 in the summer of 1960, computation time was cut to about one hour per eight hours of prediction. The 7090 was one of the world’s first largely transistorized computers.

In April, two satellites underwent environmental tests in a thermal- vacuum chamber and vibration tests at the Naval Weapons Laboratory to see whether they would be able to withstand the temperature changes in space and the vibrations and accelerations of launch. Then the two satel­lites were placed outside on a grassy knoll while the radio frequency links and solar cells were tested.

In those early days, APL had no clean rooms to prevent contamina­tion of the spacecraft, and Commander Pollow recalls asking someone to stop flicking cigarette ash on the satellite. The clean rooms that were then creeping into industry were not clean rooms as we know them today. They were intended, say the Transit members, to provide a psychological envi­ronment that made people more careful. APL’s approach to instilling greater care relied on Kershner’s cradle-to-grave policy, which was a good way of encouraging the design engineer to satisfy the test engineer.

Besides testing, April 1959 saw work begin on three permanent and two mobile tracking stations. The spring schedule called for complete installation and checkout of ground stations by June 1; the completion of thermal, vacuum, and mechanical tests on three satellites by July 1; check­out of the entire Transit tracking communication and computing system by July 15; and delivery of three completed satellites to Cape Canaveral by July 19. The scheduled launch date for the first satellite was still August 24.

By the end of July, the timetable drawn up in spring had slipped a lit­tle, but not by many days, nor is it clear from the progress reports why. According to “Kershner’s roulette,” a term that does not appear in the pro­gram’s official progress reports but was common currency among Transit team members, any delay was supposed to be because of some other part of the complicated process of launching a satellite, never because of Tran­sit. Whether or not the Transit team won that particular spin of Kershner’s wheel, the launch date had by July been set back to September 17.

Through the summer of 1959, Guier, Newton, Black, and Hook were putting the finishing touches to the software. Kershner made infor­mal arrangements for the Physics Research Laboratory of the Army Signal Corps to participate in Transit and to evaluate the absorptivity and emis – sivity of paints in an attempt to find the best materials for controlling satel­lite temperature in orbit. He and Gibson were in constant contact with the Cape and with the Air Force, which was responsible for the launch. Kersh – ner was discussing tracking and orbital requirements, drawing up the checkout procedures for the launch site, and visiting candidate tracking sites. The sites were required to be free from interference from nearby sources of radio waves, and the surrounding topography had to provide clear lines of sight to the horizon, or at least to the lowest elevation from which the satellite could provide usable Doppler curves. Each site’s posi­tion had to be accurately surveyed.

Finally, toward the end of July, things started to come together. There were only days to go before the satellites were to be shipped to the Cape. In the year since Guier and Weiffenbach had learned that they had funding for an experimental navigation satellite, the Transit team had worked impossibly hard. They were exhausted and exhilarated.

On July 31, the satellite in which they had most confidence was undergoing a dynamic balancing test. While the satellite was spinning at one hundred revolutions per minute, a structural support member in the test equipment failed. The satellite fell and was shattered.

Kershner spoke into the silence that followed, without anger or histrionics. He asked team members to pick up the pieces and to deter­mine what was still in working order. Somehow, despite the sick feeling in the pits of their stomachs, Kershner’s demeanor inspired them to work flat out to prepare their remaining two satellites for shipping. More than thirty years later, when Weiffenbach recalled Kershner’s calmness and the team’s response, there were tears in his eyes. Sooner or later, each Transit team member tells this story. It never seems very dramatic, but it clearly symbol­izes why they respect Kershner.

On August 2, two, rather than three, NAV-1 satellites were delivered to the Cape.

Throughout August, the tracking stations at APL, in Texas, New Mexico, Washington State, Newfoundland, and England followed a rigor­ous practice schedule. APL staff went to England in the middle of the month. Negotiations were underway with AT&T to ensure that enough teletypewriter and telephone links were in place to handle communica­tions between APL, the Cape, and the tracking sites. During the launch,

Weiffenbach found himself repeatedly reassuring an English operator that the expensive transatlantic telephone line should stay open even though the operator could hear no one speaking.

At APL’s own tracking station, Henry Elliott was in place as the chief operator. During the last few weeks before the launch, Elliott encountered a number of irritants. Phase shifters arrived on September 8 but were not working. The next day, he got one working in time for a practice alert on September 10. During the day a twenty-four-hour clock arrived. On Sep­tember 14, three days before launch, there was a dry run. On September 16, there was another dry run, following a timetable mimicking the next day’s launch. One of the two frequency pairs that NAV-1 carried to inves­tigate whether one pair alone would be enough to correct for ionospheric refraction was swamped by a nearby signal. If it had been the real thing, there would have been no Doppler recording of one of the frequency pairs.

September 17, 1959, was the day of the launch. A final and complete check of the station began at six in the morning. A tracking filter failed and was replaced. News came through that the launch was being held for an hour and a half. At 9:45 a. m. APL’s test transmitter was switched off in preparation for the launch. An announcement was made over loudspeakers asking everyone in the lab to turn off anything that could generate a radio signal that might interfere with the satellite’s signal. At last, at 10:33:27 Eastern Daylight Time, the rocket lifted off.

It was eighteen months since McClure had told Guier and Weiffen­bach of his idea. In that time, the Transit team had learned how to design satellites and had designed and built facilities for mechanical, thermal, and vacuum tests, ground stations, and test equipment for their ground sta­tions. They had carried out innumerable theoretical analyses, had written an orbital determination program, and had built and tested three satellites. They had seen their best work destroyed by a last-minute accident and had shipped two satellites to the Cape. Now the Nav-1 satellite that would become Transit- і A was climbing through the atmosphere.