The Perfect Machine

by Ronald Florence

  McCauley’s estimate of one hundred thousand dollars for a replacement disk was arbitrary. He had already dismissed two procedures for avoiding contamination in the glass as impractical. A melting tank large enough not to show a significant drop in level from the removal of the glass for a two-hundred-inch disk would cost more than one-third of a million dollars to construct. And pouring the disk in layers, over a period of days or even weeks, would require that the mold withstand the high temperatures for the entire period; it wasn’t clear that any refractory brick and any technique for cementing the complex surfaces of a mold could withstand that heating.

  Those options were so unpromising that McCauley decided that all future disks would be made by an entirely different process than any they had tried. They would use only glass that had been in contact with unused refractories. A newly lined tank would be filled with the glass mix, heated, and held long enough to fine the glass. The tank would then be cooled, and the glass would be mined from the tank in blocks, with cleanly fractured faces. To make mirror disks, the blocks of pure glass would then be placed on a mold under the beehive oven and heated until the glass sagged into the mold. They had already tried the process with smaller disks. No one knew if it was possible to sag the glass for a two-hundred-inch ribbed disk into a mold.

  McCauley could make no assurances about a delivery date. With enough glass on hand for the back orders of telescope disks, and with a growth of orders for commercial products in 1936, the 3A melting tank, which had been used to melt Pyrex for the telescope disks, had been converted back to use for baby bottles. Much of the equipment for the casting and annealing ovens had been dismantled, and many of the experienced personnel dispersed to other projects. Even if they succeeded in producing a mirror by the new process, it would require a year of annealing—with all the perils that entailed—and another journey across the country.

  The negotiations went slowly. McCauley’s analysis suggested that the glass that had been poured into the mold first should be better than the glass added later. Even in the heated casting igloo, the viscosity of the molten Pyrex permitted only limited homogenization of the glass in the mold, so the contaminants that affected the later pours of the ladles might not have reached down into the layer of glass close to the mold. If he was right, Anderson and Brown should start seeing a rapid decline in the frequency and gravity of the fractures.

  All any of them could do was wait and see.

  25

  Big Machines

  The Westinghouse South Philadelphia Works was an enormous factory, acres of space with machine tools as large as any in the world. Their specialty was machining hydroelectric equipment and steam turbines for ships, but American shipbuilding had been slack for years, the union movement of the 1930s had made inroads at the South Philadelphia Works, and the large work force meant a very expensive payroll for Westinghouse. The company already had slack capacity and what public relations firms would later call an “image” problem.

  Fortune magazine wrote that Westinghouse wore “seven-league boots” and straddled the “entire market.” Its irons pressed gowns in penthouse apartments and red flannels in dude ranch laundries, its generators provided the current to run chippers in Canadian pulp mills and air drills in Arizona copper mines, its motors drove steel slabbing mills and “one-mouse-power” electric razors. But so little of its production was products that the public had seen that most Americans had heard of Westinghouse only in association with broadcasting or home appliances. The contract to build the mounting for what the newspapers now routinely called “the greatest scientific machine in the world” would not only keep machinists and machines busy but would provide marvelous opportunities for publicity that would promote a corporate identity.

  Eager as it was for work, Westinghouse was also a company and a shop with a long tradition of doing work its way. The engineers were accustomed to working from large-scale drawings that covered every detail of the proposed work. They would then fabricate, weld, machine, and heat-treat the components, preassemble the turbine on the shop floor to make sure every part fit as intended, and finally disassemble the parts for transport to the shipyard, where they could be reassembled by shipwrights and mechanics. It was a solid, reliable procedure. Since even the largest turbines could easily be accommodated in the factory and foundry buildings, and the workers were accustomed to the procedures involved in assembling the machinery, the extra cost for the conservative approach was modest. Jess Ormondroyd, in charge of the experimental division and with many years’ experience on large machinery behind him, was convinced that the procedure was the only safe and reliable way to build large fabrications.

  The Caltech engineers had a different idea.

  As early as January 1935, when many features of the design of the telescope were still undecided, McDowell heard that Corning had received orders for one-tenth-size models of the telescope mirror—disks with the same ribbed-back design as the big mirror, but only twenty inches in diameter. He proposed that Caltech order one of the smaller mirrors and use it to build an exact one-tenth-size model of the telescope, with every system used on the big one. Not only could the model pinpoint design and engineering problems, but it could ultimately serve as a guide scope for the bigger telescope.

  The astronomers patiently explained to the navy captain that his idea wouldn’t work. The guiding mechanism for the two-hundred-inch telescope, like those on the sixty-and one-hundred-inch telescopes, would have to be internal to the telescope, relying on the main optics. Even if an external guide scope were needed, a. twenty-inch f/3.3 telescope would not make a good guide scope.

  The guide scope idea was abandoned, but McDowell was still intrigued with the possibilities of building a model of the telescope. As the design stages progressed, he had Russell Porter build a celluloid model of the telescope at one-fiftieth scale. By then Corning had begun producing the twenty-inch-diameter replica disks; one arrived at Pasadena for Caltech; and McDowell again took up the model idea, arguing that systems like the oil bearings and the spoked declination-bearing supports could all be tested on the model.

  The engineers pointed out that a model wouldn’t really test the design, because many of the engineering concepts couldn’t be scaled. The loads of a one-tenth-scale model were so much smaller, and the harmonic frequencies of the parts so much higher, that deformations, friction of moving parts, or vibrations of the model weren’t a useful indication of what would happen with the full-size machine. McDowell was quick to concede their points, but a model, he argued, would test how the various assemblies mated with one another and would provide a working demonstration of the new ideas, like the horseshoe bearing and the oil pads. He ordered the machine shop to build the model. Except that there would be no observers cage—even Caltech freshmen weren’t that small—and that the hydraulic piping ran in external pipes instead of lines within the mounting, the small telescope was meant to be an exact working scale model.*

  For some problems the model provided valuable data. Although the deflections of the tube in the model would be much smaller than in the actual telescope, the engineers put together a system of mirrors that could accurately measure deflections of 1/100,000 inch, fine enough to prove that Mark Serrurier’s tube design would work and that the complex forces in the yoke assembly would cancel one another out.

  Westinghouse built its own model of the mounting out of celluloid, using small brass weights to load the model while micrometers and electrical microswitches and microammeters measured the deflection of various components. Stress lines would also show up in the celluloid.

  McDowell originally wanted to solicit competitive bids for the mounting, the way he had done with navy projects. The bid from Westinghouse was a minimum of $800,000 and a guaranteed maximum of $1,100,000, far more than the Observatory Council could afford. Hale urged that instead McDowell negotiate a cost-plus arrangement, essentially the same sort of agreement the council had with Corning Glass. The new agreement held the cost d
own, but the bills from Westinghouse still raised questions.

  George Hale, eagle-eyed despite his increasingly frequent attacks, was the first to notice the problem. In January 1936 he read through the correspondence with Westinghouse and noticed what seemed like enormous “miscellaneous” expenditures, including revamping the foundry building ($81,250) and building a new turning rig ($62,500). There were also large expenditures proposed for moving sections of the mounting to the Sun Shipbuilding Company for annealing in their furnaces, and to other Westinghouse plants in Pittsburgh for portions of the machining too large to be undertaken in South Philadelphia. Hale was worried about the expenditures. Between the work in the optics shop and the machine shop, the grading and site work on the mountain, and the proposed contracts at Westinghouse, the expenditures for 1936 were rapidly sliding toward $600,000—one-tenth of the entire grant.

  McDowell had pushed for Westinghouse to build the large components of the mount because they had the facilities to fabricate, weld, and machine very large structures. Why, then, did Caltech have to pay to expand the Westinghouse foundry building or build a new turning rig to machine parts of the assembly?

  The answer was that Westinghouse planned to preassemble the entire structure in its South Philadelphia plant before it was shipped to California. Just the tube of the telescope was so large that to assemble it vertically would require excavation of the floor of the foundry, the tallest portion of the plant. Assembly of the huge horseshoe and testing of the oil pads would require major modifications to the plant. As the negotiations went on, the Westinghouse metallurgist, Norman Mochel, urged that the annealing all take place at the South Philadelphia plant, where the exact positioning of the various components during the annealing could be controlled. If fabrications were heat-treated in the wrong positions, heat-induced sagging could spoil the alignment so that components designed for precision fits would not match. To anneal the parts in the plant, a temporary furnace would have to be built to accommodate the large components.

  Mark Serrurier, sent east to investigate, thought the Westinghouse estimates were based on “meager information and very little detailed study.” Sherburne, the machinist in charge of the astrophysics machine shop at Caltech, agreed. Between the faith of the Caltech engineers in their design, and McDowell’s trust in his one-tenth-scale model, Pasadena was confident that the mounting did not need to be test-assembled at Westinghouse. The savings in money and time would be considerable. The Westinghouse engineers were equally stubborn and willing to budge on the billing if their procedures were followed. As Guenther Froebel argued, “a maximum investment of fifty to seventy-five thousand dollars now may assure the complete success of an undertaking which has to live through the years.” McDowell replied that he was sure that Westinghouse would have preferred to erect and test the huge generators they built for the Boulder Dam at the plant before they were shipped, but just as they could only be assembled and tested on-site, so the telescope could really only be assembled at Palomar. Froebel answered that the Boulder Dam generators had been preassembled for testing in the shop.

  The negotiations went on and on, and around and around. The Palomar side—McDowell by mail, Serrurier and Sherburne in person, Pease and the others agreeing from Pasadena—were eager to save money and time and confident that their own engineers and staff could put the telescope together. The Westinghouse officials were equally confident that their experience with large welded structures, not the upstart confidence of the Caltech engineers and scientists, should be trusted. Their name would be on the mounting, their publicity officials had already prepared a campaign to attract attention to the Westinghouse contribution to the project, and the last thing they wanted was a glitch and news articles about problems with the Westinghouse-built mounting.

  The customer is always right. The Westinghouse officials ultimately came around to a compromise: the tube of the telescope—Serrurier’s trusses and Kroon’s spoked declination bearings—would be assembled in South Philadelphia, to make certain the boring for the declination bearings was exactly true. The rest of the mounting, including the huge horseshoe bearing, would be fabricated in pieces, predrilled for alignment pins and bolts, and shipped to Palomar for assembly. Although Westinghouse publicity touted the telescope mounting as an “all-welded” structure, a description that sounded modern and advanced, welding the assemblies together on-site was not an option for a precision device like the telescope because the heat from welding could introduce distortions or strains. The sections would be bolted together on the site. Frank Fredericks, a Caltech engineer, would be at Westinghouse during the fabrication, and then at Palomar during the assembly, to make sure it all went together as planned.

  Sandy McDowell urged the Westinghouse officials to hold down their publicity about the project. “I know that what Dr. Hale worries about,” McDowell wrote, “is misleading or premature newspaper discussion of the various parts of his project.”

  Like almost every other company that had associated itself with the now-famous telescope, Westinghouse was eager for a part of the glory that seemed to travel with the triumph of technology. The Westinghouse officials commissioned a series of papers on the process of building the mounting, and arranged talks for selected representatives of Westinghouse at sites like Harvard University. Edward Pendray, the director of publicity at the main Westinghouse offices in Pittsburgh, issued a series of press releases about the project, touting the magnitude of the fabrications required, which would only be “handled by a company possessing the necessary equipment and facilities,” and the accuracy required for the assemblies, which, though exacting, was “nothing beyond the usual accuracy of Westinghouse practice.”

  The most demanding tolerance in the fabrication of the tube and mount was the alignment of the declination axis of the telescope tube. The tube of the telescope, in its final configuration, would be twenty-two feet in diameter, fifty feet long, and would weigh 150,000 pounds. The most critical dimension, measuring for boring the declination axis, demanded a measurement accurate to .077 inches over a length of twenty-six feet—a far cry from the accuracy demanded in the optics shop or even the .001-inch accuracy that is more-or-less routine in machine shops.

  Yet, what made the mounting fabrication exacting was the sheer scale of the pieces and the uniqueness of the project. By Westinghouse standards, the fabrications were lightweight. The largest unit they had ever built previously in the plant, a turbo generator installation, comprised a huge condenser that weighed twenty-two pounds per cubic foot, and a generator that weighed fifty pounds per cubic foot. The tube of the telescope, built up of hollow box or tubular sections, weighed only eight pounds per cubic foot.

  In many ways the telescope was a relatively easy fabrication project. It was large—the largest structure ever machined—but there was no need for superlightweight design or materials like those used in the manufacture of aircraft. Ordinary mild carbon steel was fine for the telescope, although they did take the precaution of ordering all the steel for each component of the telescope from a single heat batch, to guarantee uniformity of composition. The chief difficulty was finding or building machines large enough to handle the huge sections and devising means to accurately measure and align the sections. Each outer band of the great horseshoe bearing required a piece of plate steel four and one-half inches thick, five feet wide, and forty-seven feet long. Bethlehem Steel used a 12,000-ton forging press to form these sections to the desired curve. To align components prior to welding, and particularly while machining, the factory used surveyor’s transits on the floor of the shop. All the welding, except for the circumferences of the hub and spokes of the declination bearings, was done by hand. Rein Kroon’s design of the interior of the horseshoe had left enough room for the welders to get to every seam.

  The foundry at South Philadelphia was one of the tallest industrial buildings in the world, but even with low-slung rail cars to move the telescope sections, the tracks leading into the building had to be
lowered five feet so the sections of the mounting would clear the doorway. And even the large furnace Westinghouse built at the South Philadelphia Works would only anneal the fabrications in sections. The cage that formed the top of the telescope tube, twenty-two feet in diameter and twelve feet high, would just fit into the furnace. The biggest machine tools at the South Philadelphia plant couldn’t do the finish machining on the horseshoe, the largest journal bearing in the world. The complete horseshoe bearing would weigh 400,000 pounds, and even the huge floor mill at the Westinghouse East Pittsburgh Plant had to be extended with supplementary rollers and tracks to handle the final machining.

  The engineering calculations, in South Philadelphia and in Pasadena, went on as the work was in progress. When the final drawings and calculations were finished for the horseshoe, it was obvious that even with the superb internal framework Rein Kroon had designed for the bearing, the horseshoe would sag out of round as it turned. The turned on its side, would become a C and the top horn of the C would sag from its own weight. Manipulating a large structure to mill a shape other than a circular arc or a straight line was a challenge. The 400,000-pound horseshoe was the largest and heaviest single piece of equipment ever handled by Westinghouse, and possibly the largest single unit ever machined. The floor mill at the East Pittsburgh plant, which had previously been used to machine the thirty-foot gates for the Boulder Dam, was the only mill large enough to machine the surface.

  The solution to milling the shape was proposed by one of the engineers, whose name, like the names of so many who worked on the telescope, is lost. His suggestion has the clean simplicity of Mark Serrurier’s trusses or Rein Kroon’s bicycle-spoke mounts for the declination bearings. If the ends of the horseshoe were “pinched” and the center pushed out before it was machined, he suggested, when the tension on the horseshoe was relaxed, the shape it assumed would be exactly correct. The calculations of how much to pinch were done by Rein Kroon. To Kroon the calculations on a slide rule seemed simple and obvious; to others, including the machinists who ran the big floor mill, the numbers seemed mysterious black magic. The calculated forces—450,000 pounds pushing out in the middle and 260,000 pounds squeezing the horns in—were optimized for observations of stars that would be within forty-five degrees of the zenith. The difference in dimensions from the new procedure to boring the horseshoe unstressed was a few hundredths of an inch—the difference between the telescope maintaining a proper alignment or sagging.