Chapter 7

The Darkest Hour

The easing of Gemini’s managerial problems by mid-1963 opened the way for a concerted attack on Gemini’s technical problems. Even under new management, however, the last half of the year saw Project Gemini at its lowest ebb. The Gemini spacecraft, the Agena target vehicle, and, most seriously, the Titan II launch vehicle - each raised problems that threatened to overwhelm the program. This was to be Gemini’s darkest hour, and it began with another dual flight that raised new fears of a Soviet victory in the race for first space rendezvous. On 14 June, Lieutenant Colonel V. F. Bykovsky orbited aboard Vostok V. Cosmonaut Valentina Tereshkova followed two days later in Vostok VI. The two passed within five kilometers of each other. Once again, however, there was a crumb of hope in the Vostok’s lack of maneuvering capability. It was a faint hope.¹

Titan II in Jeopardy

Gemini’s biggest question mark in mid-1963 was the launch vehicle. Flight tests of the Titan II missile, suspended in June after two successive failures, had yet to produce results good enough to convince anyone that a booster derived from this missile was a safe bet for Gemini. To make matters worse, Brigadier General John McCoy, director of Titan programs for the Air Force Ballistic Systems Division (BSD), strongly opposed any changes in the missile to meet Gemini standards - and for sound reasons. He could not afford to risk the failure of the missile program for a chance to help Gemini.

As the Titan II program faltered, NASA concerns mounted. The Gemini Program Planning Board persisted in its efforts to resolve the impasse between NASA and BSD. On 28 June, the board asked NASA to state the least it would accept for launch vehicle performance, the Air Force to describe its program in detail. Board co-chairman Robert Seamans, NASA’s Associate Administrator, asked MSC Director Robert Gilruth for a precise statement of MSC standards for making Titan II over as the Gemini launch vehicle. The response, on 1 August, was a brief review of “Gemini Launch Vehicle Specifications and Requirements,” which pinpointed the three major problem areas that made the Titan II unsafe for manned space flight - longitudinal oscillation (Pogo), dynamic instability of the second-stage engines, and detail design faults of Titan II engines. MSC insisted “that these problems must be satisfactorily solved and the solutions incorporated into the GLV prior to its use in the manned Gemini program.”²

Every Titan II so far flown had displayed Pogo, although the level had varied, reaching a low of just over one-third the force of gravity (±0.35g) in the 17th test flight on 13 May 1963. This potential hazard to pilot safety prompted a survey of available data on human tolerance of such vibration, leading MSC to conclude that Pogo should be completely eliminated, or at least not allowed to exceed ±0.25g. A test program on the centrifuge at NASA’s Ames Research Center in California, completed in July 1963, tended to confirm the validity of this stand; an MSC astronaut test program conducted immediately after the Ames tests provided even stronger support. Higher levels might be tolerable, but 0.25g still seemed a prudent upper limit. MSC preferred an experimental program to trace Pogo to its source and eliminate it but would settle for this bearable limit if proved on Titan II flights before the vehicle flew in Gemini.³

The second major problem, combustion instability, had not yet occurred in flight, but Aerojet-General’s ground tests had revealed incipient instability during second-stage starting - that is, the initial engine-firing pulse could trigger uneven burning in stage-II engines. In a statistical sense, the engine was stable, since Aerojet-General could show that the instability rate was no more than two percent in ground tests. From a physical viewpoint, however, the engine had to be described as dynamically unstable, and that risk could not be accepted when human lives were at stake. Statistical reliability was not enough for a manned booster. Aerojet-General must develop and prove a dynamically stable engine before the first manned Gemini flight.⁴

The third major area of concern comprised a range of problems, each minor in its own right but significant in the aggregate. Of the 10 full or partial failures in the 20 Titan II test flights to date, Pogo could be blamed for only one, dynamic instability for none at all. The others resulted from small defects - a clogged injector, a failed weld, a broken line. The central problem seemed to be “a real lack of understanding on the part of Aerojet of procedures and responsiveness to problems that must be associated with the development of engines for use in a manned launch vehicle.”⁵

When several top-ranking MSC officials visited Aerojet’s Sacramento plant in July 1963, they were dismayed at what they saw and concerned about a number of questionable practices in design, manufacturing, and quality control, in general, and several components - turbine idler gears, main fuel valves, turbine seals, and turbine manifolds - in particular. The Air Force Space Systems Division (SSD), NASA’s agent for launch vehicles, had already spotted 40 engine parts that could be improved. MSC judged that most of these changes had to be made and the results confirmed in flight before the booster was committed to the first manned Gemini mission.⁶

The Gemini Program Planning Board heard NASA’s report on launch vehicle performance standards on 5 August 1963, revised the wording slightly, and accepted it. With this statement as a basis, MSC and SSD were to arrange a formal agreement on the goals of reduced Pogo, a stable second-stage engine, and improved engines. They were also to agree on the programs needed to achieve these goals and the criteria for deciding when the goals had been met.⁷

Although Titan II itself was still a question mark, the managerial logjam that had so far prevented a concerted attack on its shortcomings as a manned booster now appeared to be breaking up. Major General Ben Funk, SSD Commander, told Gilruth on 8 August that Air Force Headquarters had approved the “augmented engine improvement program.” Funk agreed that Aerojet’s efforts left something to be desired, then outlined a series of steps he had taken to tighten up the firm’s work. He had still another piece of good news. The decision to fly no more Pogo fixes on Titan flights had been reversed. The gas generator clogging problem that had marred the Titan II flight of 20 June seemed to have been solved, and the booster would soon be flying again. Missile N-25, scheduled for a September launch, would carry standpipes and accumulators to suppress Pogo.⁸

Aerojet-General began work on the improved engine program in September. That same month also saw a start on the Gemini Stability Improvement Program, or Gemsip, an effort to redesign the injector of the second-stage engine to overcome incipient combustion instability.⁹ When the Gemini Program Planning Board met again, on 6 September, MSC and SSD had agreed on the statement of “Gemini Launch Vehicle Specifications and Requirements for Major Titan II Problems” that the board had requested.¹⁰ It fully met NASA’s demands. Things seemed to be moving at last.

Titan II, however, had yet to prove itself. Missile problems had already prompted NASA, earlier in 1963, to replace one of Gemini’s manned missions with a second unmanned flight. Still unsolved, they now forced NASA to plan yet another unmanned flight. On 12 July, Mathews told MSC’s senior staff that GPO was thinking about backing up the first Gemini flight with an extra unmanned flight (making a total of 13 instead of 12) roughly midway between the first two scheduled missions, or about 1 April 1964. The proposed payload was a boilerplate capsule with instrumentation pallets like those in Spacecraft 1.¹¹

At a meeting on 5 August, the Gemini Program Planning Board agreed to review the plan. The next day, Mathews wired Walter Burke at McDonnell to begin work on the adapter that would attach capsule to launch vehicle. NASA Headquarters approved the new mission and suggested calling it Gemini 1A, or GT- 1A.* Based on data from McDonnell and SSD, the project office figured the cost of the extra flight at around $2 million.¹²

William C. Schneider, Gemini Project Manager at NASA Headquarters, presented NASA’s case for the extra flight to the planning board on 6 September. In essence, NASA wanted to guard against a failure of the first mission by planning a contingent mission, identical to GT-1, to fly before the scheduled GT-2. The board concurred, and Mathews wired Richard Dineen, SSD’s Gemini launch vehicle overseer, to make sure that the second launch vehicle would be ready in time to meet the date for GT-1A. The new mission was strictly a backup, however, to be flown only if GT-1 failed to meet its objectives. The decision waited on the outcome of the first mission.¹³

For GT-1A, MSC diverted a boilerplate spacecraft being built for flotation tests by a local Houston contractor. Named Boilerplate 1A it arrived at the Center on 24 September, where the Technical Services Division began the task of making it flightworthy. Regular biweekly panel meetings started early the next month, and the rebuilt boilerplate was ready in mid-November. It left Houston via flatbed truck on 13 December, reaching Cape Canaveral three days later, there to have its wiring and equipment installed; the work in Houston had been limited to the structure. The adapter, built and instrumented by McDonnell, arrived at the Cape 27 January 1964. By then, however, the threat that had called forth the effort had largely dissipated, and little further work was done before GT-1A was formally canceled on 17 February.¹⁴

That cancellation reflected a striking turnaround in Titan II prospects from their lowest ebb during the summer and fall of 1963. BSD resumed the flight test program on 21 August. Although the flight itself was a success, NASA suffered another setback. This missile was the first of five planned to carry the Gemini malfunction detection system, crucial for Gemini because it was to provide spacecraft pilots with the data they needed on existing or impending booster problems during launch. BSD had agreed to fly the system “piggyback” - installed, working, and reporting to ground receivers and recorders, but not otherwise acting on the missile. The system flown on 21 August suffered a short circuit 81 seconds after liftoff and provided no further data.¹⁵

Titan II’s next launch, on 23 September, did little to dispel the gloom. A guidance malfunction threw the missile out of its planned trajectory. Since the missile was guided inertially and the Gemini booster used radio guidance, this had no direct bearing on Gemini. That was small consolation, however; Pogo reached plus or minus 0.75g, very nearly the worst since the disastrous flight of Missile N-11 in December 1962.¹⁶

The heart of the matter was foot-dragging by BSD on the question of flying Gemini fixes. Once again, the planning board took a hand. It decided to replace the agreement between MSC and SSD of 6 September with a more authoritative Memorandum of Understanding between the co-chairmen of the board, Seamans of NASA and Brockway McMillan, Under Secretary of the Air Force. The board directed NASA to submit another statement of requirements for the Gemini booster and the Air Force to provide a development plan, complete with costs and schedules, for dealing with Pogo, combustion instability, and engine improvement. The board specifically asked the Air Force for a schedule of all remaining Titan II flights, with a plan for flight-testing changes to reduce or eliminate Pogo and unstable burning.¹⁷

The meeting of the board took place on 11 October 1963. Four days later, the flight-test question was finally resolved. General Bernard Schriever, a member of the board as well as commander of Air Force Systems Command, called a meeting in Los Angeles of BSD, SSD, and Titan II contractors. Schriever himself firmly supported an active program to clean up launch vehicle problems. Of special concern was whether to follow through with plans to fly Missile N-25 with oxidizer standpipes and fuel accumulators. Aerospace, backed by Space Technology Laboratories, argued strongly for the planned flight, especially since engine ground tests begun in August had confirmed fuel-line resonance as the culprit in the failure of Missile N-11 and shown that fuel accumulators would solve the problem. They carried the day, winning the crucial decision to proceed with the test flight of N-25 as planned. Funk planned to see his BSD counterpart regularly and arranged for meetings between the two project managers, Dineen and McCoy, to make sure that there was no more backsliding.¹⁸

Later events were to prove that this time the question had, indeed, been settled. Meanwhile, however, only the test flights could show that more determined management was the answer to the technological problems. Titan II was still in trouble, and the weekly status reports that Seamans was getting from the Air Force Systems Command after mid-September reflected a promising beginning but little more.¹⁹ Some thought was even being given to dropping Titan II from the Gemini project altogether. The Propulsion and Vehicle Engineering Laboratory of NASA’s Marshall Space Flight Center began to study the desperate expedient of substituting the Saturn I launch vehicle for both Titan II and Atlas.²⁰

Paraglider on the Wane

Work on the reoriented paraglider program of May 1963 got off to a quick start. Before the end of the month, North American Aviation was working out techniques for launching a tow-test vehicle from the ground. This preliminary effort, which involved first a car-towed half-scale vehicle and then one towed by helicopter, was designed to show what the paraglider would do during towing and liftoff and to work out proper towing techniques, all this to prepare for that part of the new test program in which a pilot would fly the test vehicle from an altitude of 3,000 meters to a landing. NASA’s Flight Research Center also conducted a series of tow tests, the whole effort being completed in mid-October 1963.²¹

May 1963 also saw North American begin work on the other phase of the new test program, testing the deployment sequence with the full-scale test vehicle. Since this phase of testing called for the test vehicle to land by parachute, the first step was to qualify a parachute recovery system, one standard Gemini parachute backed up by a second. North American got off to a smooth start. Two drops of a small bomblike test vehicle on 22 May and 3 June showed that the system’s two small stabilization parachutes worked. The contractor quickly began testing the full system on a boilerplate test vehicle. A minor malfunction marred the first drop on 24 June, but three good tests followed in July, with only one more needed to prove the system. What was to have been the final drop, on 30 July, brought a crucial setback. Both main and backup parachutes failed, and the boilerplate crashed.²²

The company wanted to get on to the next phase of testing and argued that the failure could be safely ignored, partly because North American believed it knew how to correct the problem partly because further tests would require a new boilerplate and mean a delay in the program. The logic was sound enough, but GPO feared that, although the immediate problem might be easily corrected, its root cause - the instability of the vehicle - might produce other, and worse, problems. GPO and North American agreed on two further drop tests. McDonnell furnished the new boilerplate, which North American, on the basis of spin-tunnel tests, modified to provide a more stable suspension system. That took time; over three months elapsed before the next drop, on 12 November 1963. Everything worked, and another test three weeks later confirmed the result; the parachute recovery system was at last qualified for full-scale vehicle deployment tests.²³

Proving the parachute system was not the only source of delay. Design engineering inspections of the full-scale test vehicle on 1 August and the tow-test vehicle on 27 September produced the normal share of required changes. Wind tunnel tests of North American’s first full-scale prototype wing at Ames Research Center in October yielded too little data and had to be repeated in early December. So it was late November before the contractor could deliver the first tow-test vehicle to Edwards Air Force Base to begin its manned program and mid-December before the two full-scale vehicles arrived.²⁴ With almost two thirds of the time available under the new contract exhausted, North American had yet to begin the major flight-testing portion of the program.

By the fall of 1963, the status of paraglider in Gemini was once more in jeopardy only partly because of North American’s troubles. The inflated frame used in the paraglider design was being challenged by advocates of what seemed to be a viable alternative - an all-flexible gliding parachute, the so-called parasail. This device offered a lift-to-drag ratio ranging from 0.9 to 1.2, lower than paraglider’s but still enough to provide worthwhile range and control. It was further handicapped by its relatively high rate of descent, which required landing rockets to cushion impact with the ground. But, overall, parasails matched conventional parachutes closely enough to promise a reasonably quick and relative cheap development of a reliable device for land landing.

The gliding parachute had, in fact, competed with the inflated-frame paraglider design back in 1961, when the choice of a land-landing technique for what was then the Mercury Mark II project was being made. Although rejected for Mark II, the concept persisted as the subject of a modest research and development program at MSC.²⁵ As paraglider faltered, parasail seemed more attractive. Project Gemini’s new manager, Charles Mathews, was more receptive to parasail - or less committed to paraglider - than James Chamberlin had been. Supported by MSC Director Gilruth, Mathews called on GPO for another look at parasail. In April 1963, after the second half-scale test vehicle had crashed but before the future of the paraglider program was decided, he asked McDonnell to study changing Gemini’s landing system from paraglider to parasail.²⁶

While McDonnell pursued its study, MSC’s Flight Operations Division and Systems Evaluation Division continued testing a parasail system and pressing for its adoption. Paraglider still had highly vocal backers, however, who denied that its problems involved anything more than sequential details that would have to be ironed out for any recovery device, even conventional parachutes. Claiming that paraglider development had been known from the first to be a hard task, they objected to dropping it after so much of the work had already been done.²⁷ The lines were drawn here they had been in 1961: Flight Operations Division and the Engineering and Development Directorate still opposed paraglider; most of the project office and the prospective pilots, supported by Flight Crew Operations, favored it.

When McDonnell finished its study early in September 1963, the issue was carried to NASA Headquarters. The company’s informed guess at the cost of a parasail and landing-rocket system for the Gemini spacecraft was $15.7 million, with a good chance to be ready for Spacecraft 7. When the parasail proposal was informally presented to NASA Headquarters on 6 September, it was rejected. Dropping paraglider on the verge of flight testing, leaving nothing to show for all the time, money, and effort already spent, was out of the question. The alternative, going ahead with parasail development as something to fall back on if paraglider failed, was ruled out for lack of funds to support both tasks at once.²⁸

Although reprieved, the paraglider program did not come through unscathed. High-level talks between MSC and NASA Headquarters produced still another reorientation of the program.* The paraglider landing system program was stripped of all other objectives, leaving as its only goal proving paraglider’s technical feasibility - which meant primarily showing that the wing could be inflated and deployed in flight to achieve a stable glide - with the accent on staying within the $16.1 million budgeted for fiscal year 1964. Until that goal had been met, there was to be no further work on a prototype system for Gemini, much less on production. Gilruth insisted on a clear understanding that paraglider might still fly on Gemini if the flight tests succeeded, that paraglider’s future in Gemini had not been foreclosed.²⁹ The implication of foreclosure was nonetheless there.

Under orders from MSC, North American ceased its efforts to keep the full-scale test vehicle fitted with the latest Gemini equipment. MSC also directed McDonnell to stop all testing related to installing the paraglider, to design parachute versions of all Gemini spacecraft, and to plan on putting paraglider in the last three, the last two, or only the last spacecraft. Nothing of paraglider was to remain in the spacecraft except the option to put everything back if the flight testing succeeded. Parachutes had, by late 1963, displaced paragliders as the planned means of recovery through the ninth mission. Paraglider landing was still listed for the last three Gemini flights, but some planners, SSD Commander Ben Funk among them, assumed paraglider would not be included in the tenth mission, either, “and probably will not be carried on any of the twelve flights.”³⁰ The very fact of paraglider’s doubtful status had already begun to close off any real chance to fly in Gemini, whether it proved itself or not.

A common feature of spacecraft development, and always a matter of concern, seems to be an innate tendency toward weight growth. Gemini was no exception. A complete paraglider landing system weighed almost 360 kilograms more than a conventional parachute recovery system. Once paraglider’s place had been questioned, that difference was seen as a bonus and was simply used up. Experiments, for instance, began to encroach on as yet unfilled space allotted to paraglider, especially after January 1964, when the Manned Space Flight Experiments Board was formed. Gemini’s planners were beginning to look on paraglider as an extra demand on the payload budget, already pushing the limits set by the booster. If paraglider were to be restored, some other mission objectives would have to give way.³¹ In other words, even if North American succeeded in showing that paraglider worked, that could no longer guarantee an attempt to fly the system in Gemini. Everything rested on the outcome of North American’s upcoming effort to deploy the wing on the full-scale test vehicle in flight; although success could not ensure a place for paraglider, failure would surely bar it.

Spacecraft Systems Become More Troublesome

Work on the systems that made up the Gemini spacecraft was moving along well in early 1963. Design had largely been completed, and developmental tests were starting.³² In some instances, this revealed unexpectedly hard problems. Three systems, in particular - fuel cell, propulsion, and escape - began to emerge as potentially critical areas. As a group, these systems called for the largest advance beyond existing technology. Each was essential to a major Gemini objective, each was new to the manned space flight program, and each resisted efforts to resolve its problems.

A major innovation in the Gemini spacecraft was the substitution of fuel cells for conventional batteries as the prime source of electrical power during flight. McDonnell had subcontracted the development of this system to General Electric (GE). By the end of 1962, GE had completed facilities at its Direct Energy Conversion Operation in West Lynn, Massachusetts, to produce fuel cells. GE had also surmounted the first serious development problem: leakage of oxygen through the cell’s ion-exchange membrane, which proved to be largely the result of mechanically induced stresses rather than an inherent design weakness.³³

Solving this problem, however, exposed another. With leakage controlled, fuel-cell test units working over longer times showed degraded performance. The cause appeared to be contamination of the membrane by metal ions from the fiber glass wicks that removed water produced by the operation of the cell. Leaks in the tubes that fed hydrogen to the cell were a second source of test failures. Both problems demanded design changes. Dacron cloth replaced fiber glass wicks, and a titanium-palladium alloy supplanted pure titanium tubing, which had proved susceptible to cracking. Slow delivery of both materials, as well as the necessary redesign, began to affect schedules. Dacron produced its own problems: the new wicks touched the membrane, drew off electrolyte, and impaired cell function. Thinner wicks were an easy answer.³⁴

The test failures, design changes, and revised production techniques combined to delay the fuel-cell program. GPO began looking for ways to increase the rate of fuel-cell production and to install fuel cells at a later point in spacecraft assembly. A visit to GE in May 1963 convinced both GPO and McDonnell that the current program was unrealistic; schedules allowed too little time for testing and failed to provide for contingencies or troubleshooting.³⁵ Throughout the spring and summer of 1963, McDonnell and GE kept juggling test and production units, trying to meet ever less tenable schedules, as slippage in the fuel-cell program mounted.³⁶ These efforts were complicated by further development problems.

The project office was far from certain that fuel cells would be ready on schedule, even when GE began shifting its main effort from engineering and development to making fuel-cell stacks on the production line.³⁷ On 27 August 1963, GPO asked McDonnell for an engineering evaluation of batteries for electrical power in Spacecraft 3, the first man-carrying ship, scheduled for October 1964; the fuel cells were to remain aboard to be used only on a test load for purposes of flight qualification. When and if proper operation was confirmed, they might then be hooked into the spacecraft main electrical system. McDonnell had a plan for dual installation of batteries and fuel cells ready within a month.³⁸ Mathews then requested a design study of substituting batteries for fuel cells in all seven spacecraft planned for two-day rendezvous missions.³⁹

NASA Headquarters also took action. George E. Mueller, NASA’s new Deputy Associate Administrator for Manned Space Flight, arranged for three senior engineers from Bell Telephone Laboratories* to visit the GE plant to assess the status of the fuel-cell program.⁴⁰ Rumors were already circulating that fuel-cell problems might force NASA to limit all Gemini missions to two days.⁴¹ GE experiments had shown that Gemini fuel cells had an operating life of 600 hours in theory, but a number of factors, among them the high operating temperatures imposed by a newly redesigned cooling system, had reduced that figure to less than 200 hours in practice.⁴² Fuel-cell problems were never conceptual. As a source of electrical power for long-term orbital missions, no one doubted that cells had a solid edge over batteries. The rub came in trying to convert that concept into hardware to meet Gemini specifications - essentially a matter of nuts and bolts, compounded to some extent by managerial shortcomings. This was clearly pointed up in the findings of the Bell experts, who toured the GE plant on 29-30 October 1963.

Their key tasks were to spot the development problems that remained and to answer two questions: Could GE solve these problems? What were the contractor’s prospects of meeting Gemini production schedules? The team pinpointed technical matters of fuel-cell structures, materials, and the like, as exemplified by uneven current distribution because of poor contact between membrane and catalyst or catalyst and rib. The Bell engineers thought that GE could solve these problems, given enough time. Whether there was time, however, was something else; the team suggested that NASA might want to think about a backup program. GE was already six months late. Despite its stated intent to make up the lost time, GE would be doing well to maintain the current schedule. The Bell recommendations, like those put forward a little later by McDonnell in a survey of possible fuel-cell changes to meet Gemini operational needs, were restricted to narrow technical considerations.⁴³

Fuel-cell production came to a halt on 26 November, as two GE task groups tried to resolve persistent engineering and manufacturing problems. Testing of the stacks on hand continued, but GE could build no new ones until a thorough study had revealed the causes of poor fuel-cell performance.⁴⁴

Still fearing that fuel cells might not be ready for Spacecraft 3, Mathews instructed Walter Burke to alter the spacecraft’s electrical system to accept either batteries or fuel cells as power sources when the spacecraft reached Cape Canaveral. By mid-December, convinced that the fuel-cell system could not be qualified in time, GPO opted to fly the first manned mission with batteries. But Spacecraft 2 would be fitted with both systems, chiefly to afford a chance to qualify the fuel-cell reactant system. The reactant supply system was a distinct development. The system, subcontracted to AiResearch, stored and fed to the cells the hydrogen and oxygen they ran on.⁴⁵

There was still little reason to believe that fuel-cell problems could be resolved even for later Gemini flights. On 20 January 1964, Mathews asked Burke to begin work on a battery-operated system for Spacecraft 4. Switching from fuel-cell to battery power for these two spacecraft cost Project Gemini almost $600,000.⁴⁶ The GE task groups having completed their intensive six-week search for the causes of the problems, a meeting was scheduled in Houston on 27 January 1964, between NASA and its contractors to review fuel-cell status and to decide what to do about it.⁴⁷

Although some missions might have to be curtailed, the Gemini spacecraft could carry men aloft without fuel cells by using conventional batteries. No such easy answer existed for the escape system. Any effort to replace it with something else would not only be difficult but far more costly. In the spring of 1963, some thought the change would be worth whatever it cost. MSC’s Flight Operations Division revived a proposal to replace ejection seats with an escape tower, the system used in Project Mercury. Doubtful that the seat could be qualified in time and skeptical of its value as an escape device in any case, chief of Flight Operations Christopher Kraft urged Gilruth to start a backup program to see, at least, if an escape tower could be used for Gemini.⁴⁸

Gemini Project Office, seconded by the astronauts and Flight Crew Operations, still believed that Gemini ejection seats could be made to work. Hard-to-solve problems were only to be expected in the development of so advanced a system.⁴⁹ Things were, in fact, starting to look up. Simulated off-the-pad ejection (Sope) tests had been suspended in the fall of 1962 until all system components were ready and the complete escape sequence, including recovery of dummy astronauts, could be demonstrated. The system had also grown more complex; it now included a device - a hybrid of BALLoon and parachUTE called a ballute - to prevent an astronaut from spinning during free fall if he had to eject from an altitude much higher than the 2,000 meters at which his personal parachute was set to deploy.⁵⁰

When Sope testing resumed on 7 February 1963, the results were disappointing from the standpoint of proving the complete escape sequence - the ballutes failed to inflate and release and the personal parachute did not deploy properly. But, in the view of Kenneth Hecht and his colleagues in GPO who were in charge of escape-system development, the test marked a real breakthrough. They had been convinced that the key problem was dynamic, the relationship between rocket-motor thrust vector and the shifting center of gravity of the seat-man combination. Analysis of the data from the test revealed that they had been overlooking a significant factor in their calculations - the tendency of the ejecting mass to tip as a result of its inertia when it left the end of the guide rails. With that factor accounted for, the key problem was solved. “The remaining technical problems,” Hecht later recalled, “were in debugging the details of a very complex design.”⁵¹

That, however, was no small order. Measures were taken to ensure that the personal parachute would deploy at the low dynamic pressure associated with off-the-pad aborts. McDonnell and Weber engineers also cleaned up the makeshift additions to seat design that had piled up in the course of development. But the complete escape sequence still had to be proved. All that took time. The new package was given its final checkout on 22 April 1963.⁵² Three weeks later, on 15 May, Sope testing was under way again, with heartening results. The last four tests in the series of 12, which had begun in July 1962, were almost flawless, only an insignificant failure of part of the test gear marring the final test, on 16 July. The development phase of pad ejection testing was now complete.⁵³

Still unfinished, however - indeed, scarcely begun - was a second series of development tests, sled-ejection tests. These were not so novel as the Sope tests, being in common use for all ejection-seat development. They simulated ejection at high dynamic pressures - as might be met in an escape during first-stage booster firing. In the Gemini tests, conducted at the Naval Ordnance Test Station in California, two ejection seats were mounted side by side in a boilerplate spacecraft carried on a rocket-propelled sled running on tracks. Known as the Supersonic Naval Ordnance Research Track, it was, obviously, called “Snort.” But the delays met in Sope tests, compounded by the reprogramming of late 1962, slowed the sled program.⁵⁴

This may have been just as well, because the test vehicle was badly damaged in its first run, on 9 November 1962. This was not an ejection-seat test. The test station needed a trial run to confirm its data on sled performance and structural soundness. It got what it wanted, but a rocket motor broke loose and smashed into the boilerplate, starting a fire. Although both boilerplate and sled needed a lot of work, GPO foresaw no delay in the sled-test program itself, since other factors had already required it to be rescheduled, leaving ample time for repairs.⁵⁵

Flawless Sope tests on 15 and 25 May 1963 showed that the new seat design was working and sled tests could begin. A dynamic dual ejection on 20 June was a success, followed by a second good run on 9 August. That turned out to be the last test in 1963. The seat system went through still another redesign, this time to provide for the automatic jettison of backboard and egress kits.⁵⁶ A more serious problem, and one that persisted, had little to do with the system itself. Testing was continuously hampered by shortages and slow delivery of parts, particularly the pyrotechnic devices that were crucial to so many of the system’s functions.** ⁵⁷

Although fuel-cell and escape systems had begun to look troublesome in 1962, the thrusters on which the Gemini spacecraft relied for attitude control and maneuvering in orbit and for control during reentry seemed at first to present no special problems. The subcontractor for both these systems, Rocketdyne Division of North American, focused its research effort on developing an engine of 111 newtons (25 pounds of thrust) able to perform within specification for five minutes of constant burning. McDonnell and Rocketdyne engineers assumed that a thruster design able to meet that standard could also sustain the pulsed, or cyclic, firing that would be called for in practice. They also thought that a working, 111-newton-thruster design need only be scaled up to meet the performance demanded of the 445-newton (100 pound-thrust) maneuvering thrusters. They were wrong on both counts.⁵⁸

Then Rocketdyne began running into trouble in steady-state thruster firing. Early tests of the small thrusters showed they tended to char through their casings and to fall off sharply in performance within little more than a minute of continuous firing. When this problem was fixed early in 1963 by a makeshift strengthening of the throat region of the thruster, which allowed it to attain a full five minutes of firing and more, Chamberlin was cautiously optimistic about having qualified units ready to be installed on time.⁵⁹

That hope suffered a setback when Rocketdyne turned to pulse testing and found that pulsing thrusters burned out their ablative liners far more quickly than identical thrusters firing continuously. Char rates - the speed with which thrust-chamber liners burn up - were one and one half times greater in pulsed firing, and thrusters were failing as their lining material was exhausted and their casings burned through. Such expedients as oxidizer to fuel ratio lowered (from 2.05:1 to 1.3:1) to reduce chamber temperatures and thus char rates, thickened ablative linings, and shortened firing times (for some thrusters) could only alleviate, not solve, the problem. In May 1963, Rocketdyne had neither completed the design of the reentry control thrusters nor fired the attitude thruster through a full pulsed duty cycle. The company had fallen three months behind schedule in delivering the thrusters and other parts of the system to McDonnell for Spacecraft 3, and development testing was equally laggard.

To make matters worse, new tests revealed that the larger maneuvering thrusters could not be simply enlarged versions of the attitude engines. Rocketdyne had, so far, done very little work on the maneuver thrusters, partly because of its focus on the smaller model and partly because it had been slow to provide test hardware and facilities. During April 1963, testing of the larger OAMS thrusters had ceased altogether. The new findings now compelled the company to reactivate that test program at once.⁶⁰

Rocketdyne made one design change after another in an effort to put together a thruster that worked, with no striking success. By July 1963, McDonnell was willing to accept a version of the attitude thruster that could not be ready until Spacecraft 5. Relaxed test requirements and less stringent performance standards - lower oxidizer to fuel ratios, shorter firing times, and reduced thrust ratings and specific impulse for all engines - helped a little, but grounds for real optimism were slight.⁶¹ As the summer of 1963 drew to a close, no small OAMS thruster had achieved a full mission duty cycle. A few larger OAMS thrusters had, but too few to be sure and with too small a margin of life beyond the duty cycle. The reentry control thrusters looked a little better, largely because of the lesser demands placed on them. They had to function only for a relatively brief time during reentry and could be expected to run dry before burning through.⁶²

Even the reentry thrusters, however, hardly inspired confidence. Stabilizing the spacecraft at subsonic speeds during the last phase of reentry, from roughly 15,000 to 3,000 meters, had been intended as one function of these motors. (The other, and more important, was to hold the spacecraft in the correct attitude for retrofire to control the angle of reentry and thus to prevent either too steep or too shallow a flight back into Earth’s atmosphere.) But, in September 1963, GPO decided that the thruster problems were severe enough to warrant seeking another way to steady the spacecraft. Since the first six Gemini spacecraft were then slated for parachute recovery, GPO decided to add a drogue parachute to the system for this purpose. Development testing of the parachute recovery system had finished in February, and qualification testing was well advanced. Mathews ordered a halt to these tests on 3 September and directed McDonnell to add the drogue. The first hope, that the new system could be ready for Spacecraft 2, did not survive a close look at the effort required. It was slated instead for Spacecraft 3, the first manned spacecraft; Spacecraft 2 would fly with the non-drogue version.⁶³

Rocketdyne, still struggling to meet the 232.5 seconds of pulse operation required of the small attitude thrusters and the 288.5 seconds demanded on the larger maneuvering thrusters, received a jolt in October 1963 from a McDonnell warning that thruster life would have to be doubled or tripled. Astronauts flying simulated missions used the thrusters even more strenuously than they were designed for, and there seemed to be no choice but to widen the margin of performance. Several months elapsed before the new demands were settled at 557 seconds of pulse operation for the small thrusters and 757 seconds for the larger ones. In the meantime, however, thruster testing at Rocketdyne ground to a halt, and the program threatened to founder. No end to development testing was yet in sight, and the start of qualification testing was a long way off. During November and December, Rocketdyne undertook an intense study of the basic features of small ablative rocket engines; McDonnell began work on an alternative design, cooled by radiation rather than ablation; and GPO was thinking seriously about the drastic step of starting qualification tests before development tests were completed.⁶⁴

A New Headache

Despite its key role in Gemini, the Agena target vehicle had received far less attention from GPO during 1962 and early 1963 than other parts of the program, chiefly because time seemed more than ample. Since it was not scheduled into the flight program until the fifth mission, Agena started with seven months more lead time than the spacecraft and Titan II, and that margin more than doubled as a result of the reprogramming crisis of late 1962 and the revised flight schedule of April 1963. By the spring of 1963, although still slated for the fifth mission, Agena’s maiden flight was not expected until April 1965, 13 months later than originally planned and trailing the first Gemini mission by almost a year and a half.⁶⁵

That was just as well, because Agena development had moved very slowly. Agena’s two propulsion systems, primary and secondary, were subcontracted to Bell Aerosystems Company in Buffalo, New York. The primary system was built around the Bell Model 8247 engine, into which were pumped storable, hypergolic propellants: unsymmetrical dimethyl hydrazine as fuel, inhibited red fuming nitric acid as oxidizer. Its rated thrust was 71,000 newtons (16,000 pounds), and it helped push Agena into orbit (the main boost coming from the Atlas launch vehicle) as well as powering later orbital changes.

The major change in the new engine from the standard model on which it was based was in the starting system. Solid-propellant charges, or “starter cans,” in the standard model fed high-speed gas to start the turbine which pumped propellants to the engine. Since these cans could not be reused, the number of times the engine could be restarted was limited by the supply of extra starter cans that could be carried. Gemini required an engine that could start at least five times, and Bell proposed to meet this demand by switching to a liquid propellant starting system. Liquids were stored in rechargeable pressurized tanks, which fed them to a gas generator where they were converted to gas and transmitted to the turbine. MSC approved the change in September 1962.⁶⁶

Like the primary system, the secondary propulsion system was a modification of a system already in use. Several Agenas had carried an auxiliary propulsion system to permit small adjustments of orbits. Two major changes set off the new model, 8250, from the former system: the new secondary propulsion system was modularized instead of having its parts scattered at various sites in the vehicle, and stainless steel bellows were used in place of Teflon bladders to expel propellants from their storage tanks. The Gemini-Agena secondary system comprised two identical modules, separately mounted but fired in unison. Each module was self-contained, with propellants, pressurized nitrogen to operate the bellows, controls, plumbing, and two thrusters. The larger of the two thrusters, rated at 890 newtons (200 pounds), was intended chiefly for minor orbital adjustments, and the smaller 71-newton (16-pound) thruster for orienting the Agena just before the primary propulsion system fired. MSC had approved the modified secondary propulsion system in August 1962.⁶⁷

Bell had just started its test program when, in the fall of 1962, Gemini’s budget crisis struck. While Agena’s role in Gemini was under fire, development stopped. But when the smoke lifted, Agena was still very much a part of the program. Contract negotiations between SSD, as NASA’s agent, and Lockheed Missiles& Space Company, the prime contractor, began in January 1963.⁶⁸ Testing of Agena propulsion systems could now begin. When it did, Gemini confronted a major new problem area.

By April 1963, Bell had completed a development version of the primary propulsion system, test-fired it, and shipped it to the Arnold Engineering Development Center (an Air Force test facility in Tullahoma, Tennessee) for a series of tests to prove that the engine would restart at the pressures and temperatures it would meet in Earth orbit. Tests began on 3 May and continued over the next two months with few surprises, although two problems did emerge. One involved the turbine, which tended to spin too fast. The other trouble spot was the latch-type gas generator valve that controlled the flow of propellants from the start tanks to the gas generator. These valves sometimes opened when they should have stayed closed, failed to open on command, or stuck open. SSD reported to MSC’s Atlas/Agena panel that both problems were being closely studied.⁶⁹

Bad luck rocked the program on 15 July, however, when the two problems combined. The valve failed during a test, calling for an emergency shutdown of the engine. A mistake in the choice of shutdown procedures spun the turbine out of control and destroyed the turbopump assembly. That was the end of testing at Tullahoma. Bell planned to finish the series in its own plant in Buffalo, once the problems had been corrected.

The turbine was fairly easy to fix by adding an electronic circuit to monitor its speed and shut it down automatically if it started spinning too fast.⁷⁰ But the gas generator valve was not so simply fixed. The failure on 15 July was not its first. A new design was clearly called for. Bell set out to improve its latch-type valve, but how good even an improved version could be was a real question. Bell also went to work on an alternative design, solenoid operated rather than latch-type. Tests over the next few months lent weight to the view that a solenoid valve was not only inherently more reliable but also reduced the complexity of the engine as a whole.⁷¹

These advantages, and the still unanswered questions about the latch-type valve, swayed a meeting at the Bell plant on 15 November. The participants decided to switch to solenoid gas generator valves in the Gemini-Agena primary propulsion system and forget about latch-type valves. But development had been much delayed. Preliminary flight-rating tests had been scheduled to begin in September. Switching to the new valves would cost four months and postpone the start of these tests until January 1964.⁷²

Problems and delays also cost money. Negotiations in January and February of 1963 had set the price (including Bell’s fee) of primary system development at $4,771,030. The price tag for solving the turbine problem would be about $300,000. Total costs kept going up, especially after the valve design proved hard to resolve. Toward the end of August, the money actually being spent began to exceed that predicted. By late October, Bell’s guess at the cost of completing the program had climbed to $6.177 million, which Lockheed thought was at least $300,000 too low.⁷³

Agena’s secondary propulsion system developed along the same lines. The new stainless steel bellows produced delays and rising costs. Negotiated cost and fee was $4,395,811; by the time that figure was settled in May, Bell was already asking for an additional $500,000 for the bellows. Scarcely a month later, actual spending was passing predicted expenses as bellows and tanks required still further design work and more testing. In mid-October, Bell’s best estimate for the secondary system was $4.63 million, while Lockheed forecast $5.2 million.⁷⁴

Growing engine costs were only part of a trend that brought the Gemini-Agena program to another critical pass in the late summer and fall of 1963. Other program costs were also rising, and the comfortable schedule cushion with which Agena had emerged in the revised program of April had eroded. Shortly before NASA Headquarters sanctioned the revised program, Lockheed estimated the cost for its work at roughly $50.4 million, with $17 million needed for fiscal year 1964.⁷⁵ After meetings in May and June to settle details of the new schedule, Lockheed reported its projected total cost as $53.285 million, but SSD had set its sights even higher. NASA’s Air Force agents wanted $37.2 million in fiscal-year 1964 funds for Atlas-Agena, with $26 million of that earmarked for Lockheed’s Agena contract. GPO protested. Mathews thought that was too much money in view of the stretched-out schedule and wondered if the program could be completed at any reasonable cost with money being spent at that rate. He warned SSD that such spending could not be allowed.⁷⁶ When SSD replied on 10 September 1963, current demands were down but the price of the total program was up again, to $57.46 million for Agena and $103.555 million for the entire Atlas-Agena program.⁷⁷

As costs rose, schedules slipped. One source of delay was attempted improvements. The first Agena D programmed for Gemini was AD-13. Meanwhile, however, the Air Force had started a program to improve the standard Agena, the first of which was to be the AD-62 model. The improved version, unlike the earlier model, came equipped with Bell’s 8247 engine, which Gemini needed anyway. Since there seemed plenty of time, Lockheed’s contract was amended to replace AD-13 with AD-62 as the first Agena for Gemini, at a cost of two months. Another month or more vanished when the Air Force decided to put the restartable Bell engine in AD-71, rather than AD-62, and GPO agreed to take that one. Work on test facilities at Lockheed was slower than expected, adding to the slippage, and development problems in the propulsion systems threatened to delay the program still further.⁷⁸

The Gemini Project Office was less than happy with the course of events, its manager least of all. Mathews was concerned about rising costs, of course, but he was just as concerned with the dearth of information that was reaching him through the filter of SSD. With the Air Force running the Gemini-Agena development program for NASA, Mathews could only plead with his agent to exert more control. Not only was GPO being bypassed in the process that approved changes Lockheed wanted to make, but the project office was not always even told what these changes were. Mathews observed, with good reason, that such decisions as switching from AD-13 to AD-62 (and later AD-71) for the first Gemini-Agena were bound to cause program delays. He urged SSD to think twice about any further changes “considering the deleterious effects that improvements can have.”⁷⁹

SSD, however, was not really much better informed than GPO about Lockheed’s changes. Mathews’ protests about the lax and shallow control SSD imposed on Lockheed highlighted the gulf that divided NASA from the Air Force on the administration of government contracts. The Air Force preferred to accept Lockheed’s record in filling past contracts as proof of its competence. The government was, in essence, paying for Lockheed’s expertise. Pressing for too many details of funding technology might hinder progress, cutting into the contractor’s flexibility without adding much to its prospects for doing the work. To the Air Force, NASA’s demands for detailed technical and financial data seemed at best superfluous, at worst harmful. What NASA wanted, of course, was real control of the program, and that demanded precise and thorough information. Lockheed was merely a case in point. The conflict between NASA and the Air Force over how tight a rein the government needed to exercise spanned the whole range of contract management. For NASA, it was a basic and never-ending problem.⁸⁰

In an effort to bring the Gemini-Agena program into line, Mathews dusted off and sent to Charles Wurster, SSD’s chief of Gemini-Agena engineering, a formal statement of work that dated back to July 1963. Such a document was needed, in any case, since there had been no formal work statement since Marshall Space Flight Center had left the picture. The new statement diverged most sharply from the old in the stress it laid on schedules and management. GPO insisted on tight control of all contractors, chiefly by using the system of coordination panels to keep close watch on what was going on. GPO also wanted the last word on any changes, with none to be approved until that office was satisfied that it had every piece of relevant data. So widely did NASA and Air Force viewpoints diverge that it was 18 months and 15 versions of the work statement later, in March 1965, before MSC and SSD finally agreed.⁸¹

NASA also planned to bring the Aerospace Corporation into the target vehicle program in a role analogous to that it already held in the launch vehicle program, general systems engineering and technical direction. The official end of Mercury in June 1963 had freed a number of experienced engineers for other work. Wurster suggested, and Mathews agreed, that Aerospace had something to contribute to Gemini Atlas-Agena program, especially in view of the work it had done with Mercury’s Atlas launch vehicle. Also in favor of the plan was a chance to impose a degree of technical continuity via Aerospace across all phases of Gemini being carried out under Air Force contracts.⁸²

Even if these measures worked, however, they would take time to show any effect. In the meantime, the Gemini Atlas-Agena program was in trouble, with engine development lagging badly, funding and schedules still changing for the worse without much warning. By the end of 1963, most of the time that had seemed so ample in the aftermath of the revised Gemini flight program just eight months before had vanished. The schedule for completing Agena development and for building the first target vehicle now had no slack, and any further problems threatened to delay the first rendezvous launch.⁸³

Silver Linings

The last half of 1963 witnessed Project Gemini beset by technical problems that stubbornly resisted solution. No major Gemini system - whether launch vehicle, paraglider, spacecraft, or target vehicle - could confidently be judged ready to fly. These months, in which the approved Project Development Plan of December 1961 had scheduled Gemini’s first four flights, became instead a time of troubles; even the revised schedule of April 1963, which called for a first flight before the end of that year, proved beyond reach. And as if to underscore those troubles, the Soviet Union showed that it still held the lead in the space race; 1 November 1963 saw the launch of Polet I, a new spacecraft planned “for use in manned orbital rendezvous flight.” Although unmanned, it “described complex figures in space” that shifted its first nearly circular orbit to a highly elliptical 1,437- by 343-kilometer orbit.⁸⁴

Yet, throughout these months that seem so trying in retrospect, the enthusiastic engineers and technicians, both in government and industry, sustained optimism that transcended the hard facts.⁸⁵ Part of that optimism might be chalked up to experience. The pattern of rising costs, sagging schedules, and tough problems was a familiar one at the cutting edge of aerospace technology. Then, too, although the precise nature of Gemini’s problems could not have been predicted, they did arise where they were expected - in those systems that demanded the greatest advances beyond current technology. That the escape system, for example, should be hard to develop and qualify scarcely came as a surprise. It had to meet standards far more stringent than had ever been imposed on ejection seats before, and the general nature of the problems to be met could be, and were, foreseen.⁸⁶

Initial schedules and cost estimates tend to be based on the most optimistic assumptions, the completely troublefree development of many complex systems. And these estimates depend on guesswork when new technology is involved - informed and reasoned, to be sure, but guesswork nonetheless. Rightly or wrongly, an organization like NASA assumes that Congress, the source of the money to make things go, prefers fast, cheap programs: the shorter the time and the lower the price, the better a program’s chances for support. But there is another, perhaps more weighty, reason for planning optimistically. If time and money are provided for contingencies, then they tend to be used simply because they are there. On the other hand, starting with the strictest limits and yielding further increments of time and money grudgingly may well produce the optimum achievement of the desired goal.⁸⁷ In reality, most of Gemini’s troubles in 1963 and later were the product of careful planning and design, credited to the program’s first manager, James Chamberlin, that got the project off to such a quick and promising start. This auspicious beginning encouraged NASA to move toward a more ambitious program, to push Gemini closer to its design limits. Problems that might have looked only mildly worrisome in the context of the original Gemini concept took on a more threatening guise when the margin for error had been much reduced.

For a variety of reasons, then, Gemini workers were more confident than a backward look at the difficulties may seem to warrant. But the problems were real; and their gravity should not be downgraded even though, in almost every instance, they responded finally to efforts to resolve them.