The Realities of Space Exploration
We were pioneers, and we knew it.
arsons auditorium was crowded. Everyone was eager to hear the news as it was relayed from the Cape. They knew about the delays that had accumulated during the final countdown, heard the announcement to switch off radio frequency generators at the lab. The moments before a launch are always tense. In the final seconds the tension was alleviated, as the voice from the Cape intoned, “twelve, eleven, ten, eight, whoops, seven, six, five, four, three, two, one.” The Thor-Able rocket lifted off, carrying Transit 1A aloft. They knew that Air Force radars were tracking its ascent; that engineers were calculating position, cross checking their slide – rule calculations and sending course corrections to the launch vehicle as needed. They heard the satellite’s transmitters and knew that everything was going well.
Then the transmitters stopped. For a while, no one knew what was happening. Then came the news that the third stage had failed. In all probability, it and the satellite burned up on reentry into the atmosphere somewhere over the North Atlantic, west of Ireland. Lee DuBois, one of the mechanical engineers, looked around the room. He saw the tears of disappointment on his colleagues’ faces.
The progress reports that APL sent to ARPA were as emotionless as those that described the shattering of their best satellite the previous month. The launch failure, it seemed, could be ascribed to the retro rockets on the second stage. These rockets were supposed to slow the second stage after separation of the third, so that the second stage would not interfere with the third as it coasted away prior to firing its own engine. When the retrorockets failed, the second stage bumped into the third stage, disrupting the third stage’s ignition sequence.
Transit 1A’s flight had lasted twenty-five minutes. Its electronics had survived the launch. As soon as the nose fairing that protected the satellite on liftoff had peeled away, all four frequencies were transmitted. The lab
immediately started an analysis of the telemetry, which comprised measurements of variables such as the satellite’s temperatures and solar cell voltages.
APL also had some Doppler data from the short period before the signal was lost. They were incomplete; Henry Elliott’s record shows that one signal was lost intermittently. For a while, an operator had locked unwittingly onto some other unknown signal. Signals from a TV station in Baltimore interfered with reception a few minutes later, and halfway through the pass, one of the tracking filters lost its lock. Nevertheless, APL learned enough to confirm “at least partially” that the ground stations’ design and operation worked, according to the progress report.
With the data received the computing team also made a rough correction for ionospheric refraction. Then they set themselves a theoretical problem, imagining that the Doppler data from Transit’s brief sojourn in space had, in fact, come from a satellite in orbit. They attempted their least squares fit. Though they clearly could not check the accuracy of their “orbital determination” against prediction of the satellite’s position during its next orbit, they could check their results against the Air Force’s radar data. They found the least squares fit was closer than it had been for the Sputniks and Explorer 1 and were encouraged. Thus, though the failed launch did not yield what they had hoped for and pointed to problems that needed to be addressed, the team did learn some things.
For a definitive analysis of ionospheric corrections and to begin investigating the earth’s gravitational field they needed a successful launch. The next attempt was set for April 13, 1960. By January of that year, Kershner was coordinating preparations for Transit IB’s launch and that ol Transit 2 A. Transit IB would be similar to the lost satellite, but Transit 2A, scheduled for a June launch, would test different aspects of the proposed navigation system.
Details piled on details. All over the United States, presumably in the USSR as well, teams of engineers and scientists were slowly coming to terms with the complexities of space exploration. Memos in English and Russian were written, which, if they were like Kershner’s, covered an array of newly recognized problems that now are familiar to those in the space business: nose fairing insulation, loads on structures, details about an epoxy bond, maximum satellite skin temperature at launch, radio frequency links, concerns about deflection and vibration characteristics of the launch vehicle’s second stage, and on and on.
Simultaneously, preparations were going forward for the satellites that would follow IB and 2 A in the Transit experimental series with the physics, the engineering, and computing all being developed in parallel— at a time when computing and space exploration were new.
At the ground stations, repeated preparations were made during the first three months of 1960 to track one of the Advanced Research Project Agency’s Discoverer satellites, which was carrying a Transit oscillator (ToD Soc Transit on Discoverer).Transit on Discoverer was part of a program to develop precision tracking for reconnaissance satellites, and the launch was postponed repeatedly. The postponements complicated preparations for IB, as did expansion of the Transit control center and its communication links to encompass the other agencies that were now interested in the project and its data, including NASA and the Smithsonian Astrophysical Observatory.
During the same period, Kershner lined up the Naval Ordnance Laboratory to do magnetic measurements and experiments. The Transit team was interested in fitting its satellites with magnets to stop them from spinning (de-spin, in the industry’s jargon), control their attitude, and provide stabilization.
By March, B and 2A were in the final stages of fabrication or testing. John Hamblen (who was Harry Zinc’s and Henry Elliott’s boss) decided that some discipline was needed. He had found out that flight hardware had been released before necessary electrical and environmental tests had been run. In a casually typed note he asked that in future those fabricating the satellite proceed to incorporate a component only if an engineer had first signed the test data sheet. Verbal assurances about a particular component, he wrote, would not do. Thus, casually, at APL and doubtless in many other labs, was the need for documentation recognized, documentation that now, assert many engineers and managers, has grown out of proportion to its usefulness.
The year advanced to Wednesday, April 13, 1960. That was a long day at APL. The launch was scheduled for 7:02 A. M. Eastern Standard Time. Once again Parson’s auditorium grew crowded. Probably the room looked as it does in photographs of the launch of Transit 4B. The ashtray filled to overflowing on a table crowded with papers. Gibson, Kershner, and Newton formal in dark suits, others in shirt sleeves. Gibson standing, pipe in hand. Kershner in headphones, or telephone to one ear, hand covering the other. Newton seated, twisted slightly to view over his shoulder the clock held at eight minutes to launch, frowning, as was Kershner.
For Transit IB, the countdown proceeded. The voice over the intercom from the Cape would have been saying things like, programmer starts … gyros uncaged… electrical umbilical ejects… lift off (at 7:03 A. M.). But it was not yet time for the champagne. The satellite still had to reach orbit, which it did, though barely Instead of the nominal 500 nautical mile circular orbit, IB went into an orbit with a perigee of 373 nautical miles and an apogee of 748 nautical miles. Such a result was very inaccurate by today’s standards, but more precise orbits had to wait until those designing launch vehicles were able to perfect inertial guidance controls.
Transit IB’s orbit was, however, sufficient to allow APL to begin work checking whether two frequencies would be adequate to correct for ionospheric refraction or whether a greater number would significantly improve the correction. The answer was that two seemed to be sufficient, though more remained to be done before this question was finally settled.
The immediate task on the first day was to determine an orbit, then to predict its position for the next twelve hours. Until midafternoon, there were computer problems. Then at 15:30 they determined their first orbit. The curves did not fit well, but they thought that this might be because the satellite was still spinning. Spinning ceased on April 19. On April 20, they determined another orbit from observations of fifteen passes at five locations. Again there was a poor fit. They decided this time that the problem was noise. Like Transit 1A, Transit IB carried four frequencies for the investigation of ionospheric effects. Now they turned their attention to the second frequency pair, and the fit was better.
With the data from the second frequency pair, they determined satellite position to within 150 to 200 feet from observations of a single pass over a limited region of the earth. With data for half a day from the different tracking stations they could, assuming a simplistic model for the gravitational field and uniform air drag, determine satellite position to within one nautical mile. The longer the arc, the poorer the accuracy appeared to be. Something seemed wrong. Over and over again they looked for errors in the data and software. They could find none. It was a troubling situation.
Extrapolating from a day’s observations, they then predicted the following day’s orbit. This was what it was all about, developing a way of predicting an orbit so that its coordinates could be uploaded to the Transit satellites twice a day, enabling the submarines to fix position with respect to a satellite in a known position.
The Transit team looked for the satellite at the time and location they had predicted.
And then they knew they were in trouble.
There was a discrepancy of two to three miles between prediction and observation. While this much error had been acceptable when they were first establishing Sputnik’s orbit in the fall of 1957, it was unacceptable as the basis for a navigation system. “The satellite,” recalls Guier,“was all over the sky.” Again, they thought that it was a problem with the programming. But it wasn’t. What they had suspected but had not fully recognized, and what O’Keefe had repeatedly warned Guier and Weiffen – bach about, now came to dominate the theoretical analysis of satellite motion. Earth’s gravitational field was far more complicated than anyone then knew. O’Keefe, because he knew about the perturbations in the moon’s orbit, was expecting that satellites in near-Earth orbits would show more pronounced perturbations, but even he could not have anticipated the huge variation and the complexity of the gravitational field that was to emerge.
For the position fixing accuracies they wanted to achieve, their knowledge of the gravitational forces perturbing near-Earth orbits needed to improve considerably.
There were precedents. Others had wrestled with apparently unruly satellites. Not least of these were the men within whose paradigms early satellite geodesists were working—-Johannes Kepler and Isaac Newton. Both had struggled to understand the nature of orbits as, mystics both, they sought glimpses of fundamental truths about the universe. Kepler’s focus was on the sun’s satellite Mars; Newton’s was on the earth’s moon. In his book The Great Mathematicians, Henry Westren Turnbull writes, “The Moon, for instance, that refuses to go round the Earth in an exact ellipse, but has all sorts of fanciful little excursions of her own—the Moon was very trying to Isaac Newton.”
And very trying would be the motion of satellites in near-Earth orbits to the early satellite geodesists who, with the technology to observe satellite motion in greater detail than could Kepler or Newton, noticed a veritable plethora of fanciful excursions. The forces causing these deviations from elliptical motion needed to be accounted for so that their effect on satellite motion could be quantified and thus orbital prediction improved. It turned out also that because the irregularities in the gravitational field are due to variations in the Earth’s shape and composition, scientists reaped an unexpected and abundant scientific harvest from observations of orbits. Satellite geodesy supplied, for example, some of the evidence for the theory of continental drift and thus for theories like plate tectonics.
APL was one of the early groups observing satellite motion. They were impelled by the unlikelihood that other geodesy programs would meet Transit’s needs by the time the system was scheduled to be operational, at the end of 1962.
Like other satellite geodesists around the world, the Transit team wanted to determine the “figure” of the Earth. The Earth’s figure is not the topography that we see; rather it is a surface of equal gravitational potential (a geoid) that coincides with mean sea level as it would be if the sea could stretch under the continents. This geoid looks like a contour map. It has highs and lows that represent how the gravitation potential differs at a particular geographical location from what the potential would be at that point if the earth were a water-covered, radially symmetrical rotating spheroid (an ellipsoid of revolution), not subject to the gravitational pulls that cause tides. This hypothetical surface is known as the reference ellipsoid.
It is the differences in gravitational potential between the figure of the Earth and the reference ellipsoid that geodesists study as they seek clues to the earth’s shape and structure. At first, only the deviations in motion caused by large irregularities, such as the pear-shaped Earth, were included in geoid models. Today’s models include the gravitational consequences of localized irregularities in shape or density. In the mid 1990s, the most accurate geoid maps available to civilians were of what is termed “degree and order 70.” Generally speaking, the higher a model is in degree and order, the more detailed is its description of gravitational potentials, in much the same way as a finer scaled topographical map gives greater detail about a piece of terrain. A geoid map, however, cannot be understood by analogy to an ordinary map. The gravitational potential at a given location is attributable not only to the local features, but also to the varying lengths of gravitational pull exerted by everything else. And the higher the degree and order of a geoid map, the more geologists can infer about the Earth’s structure. Geodesists aspire in the next century to satellite-based models that will be accurate to degree and order greater than 300, the goal being
to provide data that will help geophysicists to understand the earth’s geological origins and history
The road to such comprehensive understanding of our Earth opened with the launch of Sputnik 1. Prior to the advent of satellites, geoid maps showed modest highs and lows that were a result of local measurements of gravity. The force of gravity exerted on a satellite’s motion, though, includes the sum of all the gravitational anomalies resulting from every irregularity of shape and density in the Earth. Disentangling these effects and relating them back to a specific aspect of the earth’s physical nature is a little like unscrambling an egg. Nevertheless, with extensive computer modeling the job can be done.
APL produced the first American satellite geodesy map in 1960, a crude affair by comparison with those of today. Guier and Newton led this effort and found that as with orbital determination and satellite navigation, they had again provoked hostility. Their early geoid maps showed far greater highs and low than appeared in maps from presatellite days, and traditional geodesists dismissed them as amateurs.
APL continued to produce geoid maps of increasing sophistication, but much of this work was classified. Civilian scientists at places like the Smithsonian Astrophysical Observatory and the Goddard Space Flight Center soon came to dominate the field, though APL’s work filtered discreetly and obliquely along some grapevines.
The lab’s first gravitational model contained a value for the Earth’s oblateness that was more accurate than that existing pre satellites as well as a term describing the pear-shaped Earth. Shortly afterwards Robert Newton at APL and independently the Smithsonian Astrophysical Observatory made the next big discovery, which was that the Earth is not rotationally symmetric about its axis. Just as the northern and southern hemispheres are asymmetrical, so too were the eastern and western hemispheres. A number of scientists, most particularly the Soviets, had suspected that this might be true. Later on, APL optimized their geoid maps for Transit’s orbit; that is, they only unscrambled those aspects of the egg that affected polar orbits at Transit’s altitude.
The principle involved in extracting information about the Earth from satellite data is simple to explain in general terms, but very difficult to apply in practice: observe the satellite, note its departure from elliptical motion—its “fanciful excursions”—and try to find (in the computer model) what aspect of the Earth’s shape and structure, for example a par
ticular dense structure or a liquid area, would give rise to the gravitational anomalies that would cause the satellite’s observed departure from an ellipse.
More detailed gravitational models and ionospheric corrections enabled the orbital determination group to improve their knowledge of satellite position from between two and three kilometers in 1959 to a little under one hundred meters by the end of 1964. With problems like ionospheric refraction corrected for, other problems emerged. Would it be necessary, they wondered, to correct for refraction in the lower atmosphere? Such refraction was a source of bias in their data that could make the satellite appear to be about half a nautical mile away from its actual position. Helen Hopfield, whose dignified presence could reduce unruly Transit meetings to silence, tackled this problem, and APL made corrections for tropospheric refraction.
When the Transit group compensated for motion of the geographic poles from their mean position in the early 1970s, the satellite’s position was known to within twenty-seven feet. Polar motion, caused by precession of the earth’s spin axis due to the earth’s nonuniform shape and structure, changes the position of a ground station by about a hundred feet per year, thus introducing a small error into the orbital determination and prediction. The error was negligible for navigators but important to surveyors.
By the time of the first launch, APL had stopped characterizing orbits solely in terms of Kepler’s elements (as had other groups). First, because motion within a single orbit does not exactly obey Kepler’s second law—there are small deviations, and the elements are actually average values. Second, even these average values change gradually as the orbit shifts in inertial space because of the gravitational consequences of physical irregularities in the Earth. For navigation and geodesy, averages were not good enough. It was necessary to know as exact a position as possible at given times in the orbit.
So satellite position was expressed in terms of Cartesian coordinates centered on the earth’s center of mass, with one axis aligned with the earth’s spin axis and the other two lying in the Earth’s equatorial plane. The orbital prediction was made by finding the acceleration from Newton’s second law of motion—the famous F = та that is so crucial to science and engineering, where F is force, m is mass, and a is acceleration. The value of the force acting at different parts of the orbit comes from the model of gravitation; then numerical integration of the components of acceleration, a = F/m, yields position and velocity at any desired instant of time.
If it had not been for the new generation of computers, typified by the IBM 7090, this work would not have been possible. The 7090 was one of the newest and best when it was installed in August 1960. It could perform 42,000 additions and subtractions per second and 5000 multiplications and divisions per second, and it could store 32,768 words (approximately 0.03 megabytes). The 7090 was almost fully transistorized, unlike the vacuum-tube Univac.
The Univac had been badly stressed by the orbital determination program, taking eight hours for eight hours worth of prediction. The IBM 7090 could do the same job in an hour. To run the early gravitation models on the Univac, which embodied only a few of the terms representing the earth’s gravitational field, Guier would set aside three or four weekends. Had the Univac, which contained vacuum tubes with a mean time to failure of between 15 and 20 hours, been called upon to run the gravitational models that were to appear in the coming few years, it would undoubtedly have broken down. Even the 7090 would soon have to be updated as the gravitational model grew more intricate. Today Cray supercomputers run some of the largest models; desktop machines with Pentium or 486 chips can run models of degree and order 50.
During 1960, Newton, Guier, Black, Hook, and others prepared for the transition to the 7090. The programs had to be rewritten in an assembly language compatible with the 7090’s architecture. The orbital determination program occupied four or five trays of punch cards. Woe betide the person who dropped one. And drop them they did, recalls Black, with a laugh that has an edge even after thirty years.
Black and his colleagues were also learning—painfully—about software engineering, a nascent, scarcely recognized field. Black’s job was to get the orbital determination program running. He was starting with the physics developed by scientists like Newton and Guier. They generated the equations representing the physical realities, and as they understood more about what was going on they generated more equations. Black learned early to freeze the program design and fold new equations representing the physicists’ deepening understanding of the situation into the orbital determination program in an orderly fashion rather than piecemeal. That, at least, was his aim; but Black’s position between the scientists
and the programmers who wrote and tested the code was at times unenviable. He had to force agreement out of the scientists, and he fought Guier (his immediate boss), Newton, and Kershner, telling them, “You ain’t gonna change this damn thing.”
In 1962, Lee Pryor, who retired in 1995 as the last project manager of Transit, arrived at the lab. Pryor had specialized in computing while taking his degree in mathematics at Pennsylvania State University. His first three months at college were spent writing programs in anticipation of the arrival of Penn State’s first computer. Black says that Pryor was a godsend. When he arrived at APL, the lab was putting the finishing touches to the first “operational candidate” of the orbital determination program. “We just needed to get it out the door,” recalls Pryor.
In 1962, much physics and mathematics remained to occupy the Transit scientists, but the computing was moving from their purview to that of the professional programmers like Pryor who were writing code for an operational situation rather than for research. The move was necessary because, while the scientists could write programs for their own research needs, their programs, it seems, could be cumbersome and prone to breakdown in operation.
Once work on the gravitational model was well in hand, it became apparent that the effects of air drag and the pressure of radiation from the sun would have to be considered. These were dealt with in the 1970s primarily by an elegant piece of engineering invented by Daniel De Bra from Stanford University. The navigation satellites were placed inside a second satellite. The separation between their faces was tiny. Sensors on the Transit satellite detected when the inner satellite moved toward the outer surface, and tiny rockets moved the inner satellite to compensate for these forces, before they could offset the Doppler shift. An engineering solution was necessary because the time, size, and place of the forces could not be predicted.
In the mid 1960s, the failure of solar cells threatened the reliability of the operational Transit satellites. Until this problem was solved with input from Robert Fischell, the Transit satellites tended to fail within a year of launch. Once solved, some veteran satellites exceeded twenty years in operation. The Transit team also launched the first satellite with gravity – gradient stabilization, in which an extended boom encourages the satellite to align itself with the earth’s gravitational field. APL’s first—unsuccessful—attempt with this technology was on a satellite known as TRAAC,
Doppler shift due to satellite pass.
which also carried instruments to explore and characterize the Van Allen radiation belts. Ironically, the satellite failed because of ionized particles created artificially by a high-altitude nuclear explosion—as did many other satellites.
TRAAC carried a poem engraved on one of the satellite’s instruments. It was written by Thomas Bergen, of Yale University and is reprinted at the end of this chapter. Its mixture of hubris and wistfulness captures something of the atmosphere that surrounded the early work on satellites.
In the case of APL, that work led, of course, to the Navy’s Transit Navigation Satellite System. The lab built the experimental series, the prototypes, and many of the early operational satellites. For a time, Navy Avionics built some operational satellites, but the job reverted to APL when these proved unreliable. Eventually, RCA won the commercial contract for construction. More satellites were ordered than were needed,
because a problem with the solar cells that was reducing their operational life was solved after the contract was placed.
During the 1980s, under Bob Danchik’s tenure as project manager, when GPS was nipping at Transit’s heels, these satellites were finally launched. The last Transit satellite went into orbit in 1988.
Although there was always at least one operational Transit aloft and available for the submarines from 1964 onwards, the system was not declared fully operational until 1968. At that time four satellites provided global coverage. Not the instantaneous, precise three-dimensional position fix offered by the twenty-four-satellite constellation of GPS, but still, for the first time, an all-weather, global navigation system, a system developed initially for the military, but which evolved until ninety percent of its users were civilian.
In ten years, a newly perceived consequence of the Doppler effect in the three-dimensional world grew through all the stages necessary to design and engineer a space-based navigation system. The program began at a time when vacuum technology was giving way to transistors, when programs were written laboriously in assembly language, and when no one knew how to develop large software applications. The conditions in space were unknown. The physics of the newly entered environment had to be analyzed theoretically and understood experimentally. The complex nature of the earth’s gravitational field had to be researched and a provocative new understanding of the geoid developed. Launch vehicles were imprecise in their placement of satellites, if the satellites reached orbit at all. Satellite design was a new field, with stabilization, attitude control, and communication between space and Earth all unknowns.
During the early development of Transit, the launch vehicle changed, the computers changed, and programs had to be rewritten. Ground stations and satellite test facilities were built. Programs and equipment had to be developed for the submarines. It is hardly surprising that one or two people say that they ended up in the hospital, nor that the effort is remembered vividly and affectionately. But as Pryor noted shortly before he retired in 1995, it was time for the Transit program to end.
When the Navy switched off the last Transit satellite in early 1997, it ended the longest-running singly-focused space program to date. It severed the last direct link with the opening of the space age, closing the doors on that shed on the plains of Kazakhstan and on the cold morning when Sergei Korolev thanked his exhausted and elated engineers who had launched Sputnik I, which Guier and Weiffenbach would track, providing the basis for Transit, which helped Polaris, America’s riposte against the Soviet threat of nuclear attack, firing the rockets with whose development Korolev had been so involved, because he believed that rockets were defense and science, which they became, for both sides, as did Transit, which also became important to civilians. Here is one thread in the Cold War.
And one wonders.
What would have happened if McClure, say, or Pickering, Milton Rosen, or von Braun had met Sergei Korolev? If they had been in a room with chalk, blackboard, and a problem? Faintly, one hears the voices, discerns in imagination the energy and the imminent verbal explosions as Korolev’s little finger lifts toward his eyebrow….