Good afternoon, everyone! As you may recall, last week we covered the policy of the 43rd President of the United States, Ann Richards of Texas. However, we ended on a bit of cliffhanger with regards to the first Artemis Moon Walker , Don Hunt, and a new launch entrant, Star Launch Services. This week, we're picking up that thread, and backfilling a little. Hope everyone enjoys it!
Eyes Turned Skyward, Part IV: Post #4
The 1990s, like the 1980s before them, saw their own generation of spaceflight startups, fueled this time by the rising buzz about the “need” for large constellations of low-Earth-orbit communications satellites and the plethora of companies seeking to build them. Undoubtedly, the best funded of the gaggle of firms looking to cash in on what looked to be a growing launch market was Paul Allen's Star Launch Services, Inc, founded in the early 1990s after Allen was approached to invest in one of the early constellations. After studying the market, however, Allen became convinced that launch, not constellations, was where the big money was to be found, and where the big improvements would need to take place to make further progress in space development [1]. Besides his own considerable personal fortune, Allen was able to obtain support from several other Microsoft executives, including Bill Gates, in founding his new space launch company, theoretically endowing StarLaunch (as it was nearly universally known) with billions of dollars of backing upon its founding in 1994.
After securing start-up funding, Allen’s first act was to aggressively headhunt among Grumman’s Starcat veterans, now at work for Boeing. For Starcat’s chief designers, the decision by the government to not pursue their vehicle as a basis for further development, instead favoring the Lockheed Starclipper project, left them with a sense of unfinished business. Given the chance to develop the Starcat concept into an SSTO capable of a payload of several tons--enough to carry several LEO comsats to deployment orbits--many of them bolted for the new firm. This vehicle, personally named Thunderbolt by Paul Allen, would also be capable of transporting humans using a crew transport module filling the cargo bay, or supplies for Space Station Freedom or Mir, in an effort of diversify the potential customer base. With the hiring of substantial portions of the X-40 team, StarLaunch began development on Thunderbolt in earnest in 1994 at the ex-Rockwell plant in Seal Beach, California where the Saturn V’s S-II second stage had once been built.
However, it wasn’t long before the company encountered issues, primarily caused by the basic physics of the problem. To achieve single-stage-to-orbit launch performance with engines of similar performance to the J-2S or the R-10, Thunderbolt would have to achieve a structural mass less than 10% of the vehicle’s gross weight. To do so would require keeping close to the dry mass fraction of the very best hydrogen-oxygen stages ever built, despite having to design the stage more robustly for reuse and include weight-hogging recovery provisions. Moreover, only meeting this goal would not allow payload; any meaningful performance would have to come from either better engines or even more aggressive structural targets. Though Thunderbolt’s engineering team spent two years working long hours at both ends of the problem, their efforts were not enough to close the design. While the engineering team had originally predicted in 1994 that Thunderbolt would be flying by the end of the decade, by 1996 the first orbital flight seemed like it would be a decade or more away. Worse, in the meantime the initial “bubble” of companies seeking to launch constellations had begun to thin out (which included the cancellation of the massive “Teleworld” global internet system), and it began to look to Allen as though even if the problems of reusable SSTO could be conquered, it might not happen before their target constellations were either launched on more traditional vehicles or had faded away, unable to close their own business cases.
In frustration, Allen hit the reset button on StarLaunch’s development program, charging the largely ex-Grumman engineers with a complete design overhaul while refreshing the department with new engineering graduates. Rapidly, the team came up with a strategic concept that addressed the communications satellite market with a combination of technical simplicity and reusability, while building on the work already carried out. Although the existing work had run into a performance wall short of the requirements for a useful SSTO vehicle, the ex-Grumman engineers now leading the program pointed out that they could use that effort to create a substantial leap from the basic Starcat design’s capabilities, sufficient to enable a fully reusable two-stage-to-orbit system with a quite significant payload fraction. However, despite the advantages of not starting from scratch and the firm’s theoretically enormous potential capitalization, Allen was unwilling to use it as an efficient method of destroying the wealth he had earned from Microsoft’s rapid growth the previous decade, should even the revised design not pan out. Therefore, rather than leap to the final fully reusable design, the company would start off with a basic launcher, the "Thunderbolt L1" (for Launcher 1).
The fundamental concept behind the Thunderbolt L1 was developed from observations of several simple facts about launch vehicle first stages. First stages usually make up most of the cost of the launch vehicle; they have to be large and strong in order to bear the weight of the stages and payload above them, and have sufficiently powerful engines needed to lift these payloads. They also, however, face the least demanding flight regime of all of the stages, reaching speeds much lower than upper, orbital stages with correspondingly lower aerothermal loads. Finally, weight growth in a lower stage has a reduced impact on payload mass; one extra kilogram in a first stage ‘costs’ far less than one kilogram in payload, while an upper stage growing a kilogram directly subtracts from the maximum payload it can lift. Therefore, if the work already done on the SSTO was leveraged to build a reusable first stage without worrying about reusing the second stage, Thunderbolt would obtain most of the benefits of a fully reusable launch vehicle at a fraction of the engineering effort and cost. While this would require the ongoing production of expendable upper stages, the existing size of the planned SSTO Thunderbolt meshed well with the size of the flight-proven (and readily available) Centaur-E upper stage. Combined, this pairing would make up the “L1” system, which could provide a roughly 6-ton payload to orbit or just over a ton to GTO. In the future, the Centaur could be replaced with a reusable second stage to create a so-called “L2” design, which would finally achieve the company’s full-reuse goals. In the meantime, while the interim L1 would have a higher cost-per-kg than the original all-reusable plan would have allowed, it would be dramatically cheaper than the other fully-expendable launchers in the targeted size class, such as Europa 2-HE, the ALS Carrack, and some of the smaller Delta variants.
Again to minimize development, the Thunderbolt L1’s reusable lower stage would use a sea-level variant of the relatively high-performance, but still robust J-2S upper stage engine. Already extensively redesigned from the original J-2 for improved performance but decreased cost (such as the switch from a separate gas-generator to a cycle that used tapped-off combustion gas to drive the vehicle’s turbopump, resulting in lower parts count and improved ruggedness), StarLaunch had contracted with the engine’s manufacturer, Rocketdyne, in 1993 to provide a “sea-level” version of the J-2S, similar to how the RL-10-A-5 used in the Starcat had been tailored to offer better performance at the surface than its space-qualified siblings. Rocketdyne was then to demonstrate the result on the test stand for several sets of dozens of firings with minimal maintenance between firings, proving a capability for the longer-duration operations with minimal maintenance that were key to StarLaunch’s strategy. This J-2S-3 engine, suitably clustered, would offer enough thrust and efficiency for the nearly-200-ton Thunderbolt first stage even with a very conservative structural margin. After separating from the upper stage, the first stage would be recovered through the combination of aerodynamic control and vertical propulsive landing proven extensively by the Starcat team. Given that much of the design could be drawn from the previous work on the SSTO Thunderbolt concept, the plan was that the L1 could enter flight testing before the new millenium after an aggressive 6-year design, test, and production cycle.
As development began to transition to the new L1 concept at the company’s headquarters at Seal Beach, the search also got under way for one of the other key elements of the project: an operations site. In order to operate with the rates they hoped to--upwards of ten times a year for the L1, and hopefully even more often for the L2--StarLaunch would need a launch site that would not weigh them down with operations overhead from many other operators. For this reason, the Cape was right out--the sheer volume of Air Force, NASA, and Lockheed-McDonnell flights was so great that it would be difficult to fit in yet another operator, let alone one planning to fly as often as Thunderbolt. Another site would have to be found. Initially, StarLaunch had hoped to be able to build up its own operations at Matagorda, Texas; however, ALS was the main operator of the site, and was unwilling to negotiate joint operations of the site they had developed with so much time and expense with a potential rival. While Texas’ state government tried to pressure ALS into accepting StarLaunch’s offer and offered tax and other incentives to lure StarLaunch to the potentially lucrative spaceport, they also received a more attractive offer from the state of Virginia. Since before NASA was created out of NACA, Wallops Island, Virginia had played host to a sounding rocket launch site, the Wallops Flight Facility. Although a critical test site, it had never handled the kinds of monster boosters--and monster investment--that sites like the Cape and Vandenberg received, and was unknown to the general public. The government of Virginia was keen on fixing that as the commercial space age dawned, and worked with NASA Wallops to acquire land south of the existing NASA launch pads for heavier commercial vehicles. They came to StarLaunch with an attractive offer--first occupant at the new Wallops Island Commercial Spaceport (WICS), with discount rates stacked on top of a package of financial incentives. Thunderbolt operations at WICS would be able to rely on Wallop’s existing, relatively unused range infrastructure, without having to compete with other commercial operators for launch slots. It was too good an offer to refuse, and StarLaunch signed on in March 1998 as the first operator for WICS. Fortunately, like Starcat, Thunderbolt had been designed for minimal pad operations requirements, simplifying and reducing the amount of construction needed to create an operational launch facility. The addition of a second stage complicated matters and required a new integration facility, but the pad itself was relatively simple.
Meanwhile, the design of the Thunderbolt L1 was beginning to stabilize. In overall concept, it was effectively an overgrown Starcat, relying on a vertical takeoff-and-landing design with four J-2S-3 engines clustered under the first stage. Unlike Starcat’s separation of propellant tanks and aeroshell, Thunderbolt would utilize a monocoque design where tank skin and aeroshell were the same structure, saving weight. Only the fairings needed to house auxiliary systems, the attitude control thrusters, landing gear, and the vehicle’s control flaps would be structurally distinct. The booster stage would loft the expendable Centaur upper stage to roughly Mach 9 before separating, firing its engines to reverse course towards the pad, then carrying out an aerodynamically controlled glide-back prior to re-orienting tail-down for a high-acceleration final landing on any two of the four engines. While many of the key concepts had been proven by Starcat and rough technical solutions developed prior to 1996 could be used as a base, the process of designing a lightweight, robust reusable booster was intense--and the margins for economical performance were tight. As the first boilerplate booster began production in 1999, the effects of these tight margins became apparent--it was almost 2 tons, or nearly 10%, over the design goal weight of 23 tons dry weight, resulting in a payload shortfall of almost half a ton. While some of this weight was extra margin built into the prototype for testing, and would not be present on the production models, the vehicle would still need to go on a serious diet after its first trials to reach its original performance targets. Additionally, the “simple” task of adapting the Centaur to suit Thunderbolt was proving to be more complex than StarLaunch had bargained for. While Northrop was happy to supply the stages, the existing demand from other Centaur users was enough that they insisted that when problems of interface arose, Thunderbolt should adapt to meet Centaur, not the other way around.
The stress on the StarLaunch team as they worked through the design hurdles of Centaur integration and as the prototype booster stage took shape in the same working spaces where the S-II stages had once began their preparation for pushing men to the moon was intense. The financial burn rate of the project was such that more than two thirds of the startup funds had already been spent, and while Allen was very engaged in the project (though not on a day-to-day engineering level), he was quite clear: having already endured one reset, the company needed to produce results; he wasn’t willing to just perpetually fund development projects. StarLaunch’s designs needed to start proving themselves for the company to avoid going down in history as just another example of the classic truism about how to make a small fortune in the rocket business: start with a large one.
Despite this Damoclean sword, or perhaps thanks to a combination of it and the ambitious goals, the Thunderbolt team’s morale remained high, and the project made continuous progress. Finally, and the production of several test articles later, fall 1999 saw the Thunderbolt L1-Alpha undergoing final shipment preparations at Seal Beach. While the Starcat before it had been light enough to travel by air using the Super Guppy transport, the larger Thunderbolt would have barely fit in the plane’s cargo bay, and the company had instead decided on barge transport via the Panama Canal. Thus, just before Halloween of 1999, L1 Alpha was loaded onto a barge at the same port which had once handled the S-II and began the trip to Wallops for fit checks with the newly-completed pad facilities there. The time required for this process and initial static testing of the pad fixtures, combined with the rapidly-chilling weather meant that the first flights of the vehicle would have to wait for the new millenium.
Early testing of the prototype Thunderbolt was similar to that faced by the Starcat program--ground handling trials, followed by wet dress rehearsals to test pad fittings, followed by firings of the engines with the vehicle ballasted with fuel and the engines throttled to below liftoff thrust. Finally, on March 9, 2000, Thunderbolt made its first hop from the landing apron (flights from the launch mount would have to await proof the the vehicle could translate in mid-air). The first hop was nothing impressive--a mere 2 meters. However, StarLaunch’s engineers and technicians were quick in validating the data against their models with Starcat alumni depending on their experience with that program to advance Thunderbolt's flight testing, and the flights grew ever-more-aggressive throughout 2000. By the end of its first year of operations, Alpha had flown 6 times, reached an altitude of over 5 km, demonstrated mid-air control with its thruster and body flaps at subsonic and supersonic speeds, and demonstrated idling its engines to near-zero thrust and falling only to successfully “stick” landings just as its little sibling Starcat had before it. The next step was to prove the actual flight profile--the most ambitious flight yet. For this flight, an inert second stage arrived at Wallops from Northrop in March (a Centaur stage rejected near final assembly due to structural cracking which made it unacceptable for use on a proper launch) and it was mated to the booster prototype inside the integration building before being moved to the pad.
For the first time, a complete look-alike for the final Thunderbolt L1 system sat on the launch mount at Wallops in April. The launch was attended by most of the management and engineering team, including Paul Allen himself, as well as other investors and observers. The liftoff was as nominal as Thunderbolt’s controllers had come to expect--the extra weight of the ballasted second stage reduced the vehicle’s acceleration, but seemed to cause no other hassles. As the rocket’s exhaust trail piled up behind it on a tower east over the Atlantic, the vehicle’s controls seemed to be smoothly compensating for the extra mass, and the separation was completely nominal. Separation pyros on the inert second stage fired to give some space, then the “payload” of more than 25 tons of metal and water ballast plunged on in a ballistic track towards the ocean as the Thunderbolt turned back towards Wallops. However, when the vehicle’s engines once more idled for the RTLS coast, the aerodynamic controls initially failed to pick up the slack--a hydraulic failure in one of the body flaps impaired the vehicle’s yaw control. The vehicle smoothly switched into a more fuel-intense control scheme that used the attitude jets to boost yaw authority. While the flight was not as smooth as hoped, the landing was nominal other than the hit to the fuel margin. However, margin is there to be used, as the engineers noted while debriefing the flight shortly before Easter.
In post-flight analysis, the failure of the body flap’s hydraulics was found to be a control cable that had been improperly connected during maintenance between the solo test flights and the all-up test. The connector had shaken just loose enough in flight to cause intermittent cutout in the affected flap, leading to the issues. Having flown seven flights, a quarter of the vehicle’s initial design lifetime, Alpha was returned to the factory for complete inspection, inside and out. The results were heartening--the vehicle was still slightly overweight, but it had actually held up better so far than the designers had dreamed. In addition, the very public test flight sequence--and the sequence’s successes--reassured customers who had booked launch slots on Thunderbolt, not to mention potential customers who had been waiting for StarLaunch to deliver before putting their money on the line. When the flight program resumed in late 2001, this time with the second booster off the line (Thunderbolt L1-Beta), the first operational L1 flights appeared to bear this out. After its own series of commissioning hops through the last months of the year, Beta made the first full orbital flight of the program in February 2002. As expected with it pedigree, the Centaur upper stage performed absolutely nominally, placing the payload simulator within 5 km of the goal orbit. A second burn was used to raise the orbit’s apogee 90 minutes later, demonstrating the orbital maneuvering that would be key for GTO flights or for placing multiple LEO satellites into differing final orbits with one flight. By the time the second stage had burned out, the first stage had been on the ground for well over an hour, and was already beginning de-servicing and preparation for the return to the hangar for the vehicle’s first commercial flight.
Finally, nearly two years to the day after the arrival of Alpha at Wallops, Thunderbolt Beta carried its first paying payload, a LEO satellite internet trunk relay intended to eventually serve southeast Asia. For many competitors who had dismissed the possibility that StarLaunch could pull it off, the flights were disturbing. At roughly $2500/kg to LEO, the L1 was about half the cost per kilogram of the ALS Carrack, and an even smaller fraction of the slightly more expensive Europa and Delta launchers. While competition from Russian (and increasingly Chinese) expendables could approach this cost, Thunderbolt had an advantage in the US market, and to a limited degree in the European one. As the program had moved towards flight, many payloads suitable to the lifter had secured reservations on both StarLaunch and alternate providers, seeking to back up Thunderbolt’s cheaper costs with a more reliable alternative if the new rocket couldn’t make the grade. However, it seemed like the worries were unfounded--by the time Thunderbolt L1-Gamma entered the rotation in late 2003 after its own series of acceptance test flights, Star Launch Services’ launch record for its first year-and-a-half of operations was a spotless 6-for-6, with Beta demonstrating a smooth flow for turnaround between flights. The plans for 2004 would see Beta make four additional flights before its own first major service, while Gamma and Alpha were to ramp up operational tempo to launch as many as twelve flights.
However, as the company prepared to gear up to meet this goal, the market for payloads to require such a rate had begun to collapse. The technology sectors had been under increasing skepticism from investors after the rapid growth of the 90s, and the ubiquitous venture capital that had been a keystone of the comsat constellation boom was drying up, leaving firms which had already launched struggling to reach the critical mass of satellites they needed for continuous global coverage, while others withered on the vine. Worse, the build-out of terrestrial alternatives such as fiber-optics and cell phone towers were beginning to undermine the business case of many of the comsat constellations. More general economic troubles were enough to spell the downfall of the constellations. Most had not even developed and launched hardware, and quickly slid into liquidation. The few survivors were mostly those that had been tied into the US government’s satellite air traffic control system, and which were therefore less vulnerable to competition in telecommunications. Even so, many were still forced into reorganization to eliminate the huge and now unserviceable debts they had built up developing and launching their systems, leaving them ripe targets for takeover or buyouts. With the vanishing constellations went all but a tiny fraction of the plethora of entrepreneurial space firms that had grown up during the 1990s, almost all of which had relied on the constellations for their business cases and therefore lost any hope of customers with them. With the constellations gone and even geostationary deployments drastically scaled back from mid-90s predictions, the business case for StarLaunch was badly dented, and by all rights Allen could have simply written off the company at that point as a failed investment--the more limited markets available could hardly repay the initial investments quickly.
Instead, Allen looked at StarLaunch with new dedication. While the vast market StarLaunch had originally been aimed at might have dried up, the company’s Thunderbolt L1 was by any measure dominating the market section it occupied, and if the rapid attrition of payload reservations meant that meteoric growth and the funding of the L2 reusable upper stage solely through revenue no longer seemed likely, it only meant that StarLaunch had spare launch capacity with which to target other markets. To this end, Allen had already recruited Don Hunt, famous internationally as the commander of the Artemis 4 mission, and put him to work with a simple goal: identify ways that StarLaunch could leverage its existing achievements and launch manifest to continue to forge a key, profitable role in the future of spaceflight. It was a pitch Hunt was unable to resist, and he set to work immediately with the belief that NASA and StarLaunch could forge a mutually beneficial future--a vision on which the company’s continued growth would depend.
[1] Just as a general note, our inclusion of Paul Allen’s participation in this venture actually predates his Stratolaunch announcement in OTL, though by only a month or so. [2]
[2] As an even later note, our inclusion of the fin issue on the first demo mission was written back in the summer of 2013...