The Secret in Building 26

by Jim DeBrosse

  Starting in April 1942, the new guard wasted its first six months of Bombe development asking the British for many of the same technical details that had been offered to Driscoll. By May 1942, the Americans still had not adopted key parts of the British Bombe logic—including the cold-point test and the two-way flow of current in the Turing design. OP20G wouldn’t even realize the full power of the Diagonal Board until July, when its inventor, Gordon Welchman, explained to American visitors at Bletchley Park how his device could reduce the number of false hits by 60 percent.

  Yet the British can’t carry the blame alone for the delay in the American Bombe program. Even as late as mid-1942, the Americans were saying they didn’t need a copy of the British Bombe, just diagrams of Bletchley’s latest machines. They were acting as if they had their own Bletchley Park. Through the spring and early summer, the OP20G engineers continued to explore the possibilities of a high-speed electronic machine—hoping for breakthroughs in technology and counting on advice from the more experienced British. They eventually did receive the needed assistance from Bletchley, learning more about Britain’s cryptanalytic methods and how to avoid testing all the Enigma-wheel combinations.

  But the Americans didn’t want to depend on the British for the captured Enigma wheels, codebooks, and cribs to make their own Bombe method work. OP20G urged its engineers to design and build several powerful new machines that could make statistical and catalog attacks practical, without the need for crutches. The Navy still hoped to attack Shark with Wenger’s old dream of pure analysis.

  At the start of the summer of 1942, Bletchley Park hinted it had found a solution to M4 Shark. But as the summer wore on, the Navy tired of waiting for the British four-wheel Bombe to work and began laying the groundwork for their own Bombe project. By August, the Americans were determined to tackle the problem themselves—alone if they had to. OP20G informed the British that month it was going to build its own Bombe, and within several weeks it hastily announced that its men had beaten Britain and C. E. Wynn-Williams to the design of a fully electronic Bombe with a two-way electronic wheel.

  The apparent design breakthrough was aimed at the Enigma challenge alone, not at creating a general purpose codebreaking machine. The project quickly took shape: Wenger was given all the money and staff he desired plus freedom from the traditional oversight and meddling by the U.S. Navy’s Bureau of Ships. NCR’s electrical-research laboratory, headed by Joe Desch, was taken over by OP20G to be a research center and possibly a production site for the machine.

  Months before, Desch and his men at NCR had been screened for security, and began to move their laboratory from the Building 10 headquarters on Main Street to the old night school in Building 26, where they would be separate and secure from the rest of the NCR campus. In the midst of a late November snowstorm, Desch and his staff carefully stacked crate after crate of their delicate equipment onto the back of a flatbed truck. They stayed there with their load as it was driven at a crawl down three blocks of slippery streets to Building 26. Later, Marine guards with riot shotguns and submachine guns were dispatched to Dayton. A call went out to naval personnel to recruit the best electronic engineers, whether or not they were in the Navy, and Desch was ordered to expand his seventeen-man civilian staff. The M section was reorganized to oversee the electronic Bombe work at NCR. Then, in early July 1942, the Navy approved an open-ended contract that could provide, if needed, millions of dollars for the project.

  Now the pressure was on at OP20G, and not just because of what was happening in the Atlantic. The Pacific was still a disaster, even though the victory at Midway in early June signaled a turning of fortunes. That victory was something of an embarrassment for Wenger and those in OP20G who had fought so long for centralized operations and expensive machines. The turning back of the Japanese fleet was in large part due to the Navy’s long-awaited reentry into the Japanese naval-operations code, JN-25. But it was a “mustang” officer, Joseph Rochefort—an ex–enlisted man and Dayton native commanding OP20G’s tiny outpost in Hawaii—whose unit had stripped off the JN-25 additives and translated the exposed code groups, enabling the Navy to understand key parts of Admiral Yamamoto’s impending attack in the Pacific. Worse for OP20G, the Hawaiian outpost did so with only a handful of tabulators and sorters run by Navy bandmembers from the USS California, who had little else to do since their ship had been sunk during the attack on Pearl Harbor.

  The codebreakers learned that Yamamoto was out to repeat his stunning success at Pearl Harbor by concentrating his best ships, including his aircraft carriers, for a final knockout blow against the American Navy. Rochefort’s group was fairly certain that the letters AF in Yamamoto’s directive stood for the coordinates of Midway Island, but OP20G’s Washington analysts insisted it was the Aleutians.

  Rochefort devised a clever test: he had the Navy installation on Midway Island broadcast a message that its desalinators were no longer working and that its supply of freshwater was dwindling. Sure enough, within a day or two the Japanese broadcast their own message that AF was nearly out of freshwater. Admiral Chester Nimitz quickly devised a plan to ambush the Japanese at Midway, and Yamamoto’s fleet steamed into a fatal trap.

  Rochefort’s correct analysis was a blow to OP20G’s grand schemes for scientific cryptanalysis. Wenger, Engstrom, and their college men had to show results in the Atlantic if the M section was to survive. But they would soon learn the same hard lessons in humility that Shark had taught the British. The ambitious path they had chosen was to prove much longer and more tortuous than any of them could have imagined.

  5

  A Giant Leap . . . and a Step Backward

  Summer–Fall 1942—Dayton, Ohio

  FOR WEEKS IN July and August, sweltering in temperatures that hovered near ninety that summer, Joe Desch pored over plans for what promised to be a remarkable new codebreaking machine, developed by the Navy’s M section of mathematicians and engineers. Even though he had overseen thirteen other military projects at NCR, this, he knew, would be his ultimate challenge: a totally electronic deciphering machine at least one hundred times faster than anything in use. The Navy design might well rely on tens of thousands of the miniature, fast-pulsing tubes Desch himself had invented, pushed to the limits of their capabilities.

  Perhaps no man in America knew those limits so well as Desch. Prior to the war, in his laboratory on the third floor of NCR’s Building 10, he had created the fastest miniature thyratron tubes of his day. He himself had blown and shaped the tiny bulbs around their delicate cathode wires. The nickel wires first had to be superheated to fifteen hundred degrees in a quartz chamber filled with hydrogen. With the flammable gas leaking from the chamber, “you had to light both ends [of the furnace] exactly right or you had an explosion,” recalled Jack Kern, a former NCR engineer who had worked under Desch. “That was the fun part of the day.”

  The tube assemblies, each about two inches long, were baked again at five hundred degrees to drive all the residual gases out of the glass. Each tube was then injected with a prescribed amount of argon. The inert gas conducted electricity at a very specific threshold of energy, with no partial state in between. Hence, the sealed tubes were ideal “on-off” switches that could be used for counting infinitesimal pulses of electricity—more than one million per second. Desch’s tubes had been the first to reach such speeds, in a counter he had designed for the Army’s Aberdeen Proving Ground, for precisely timing the flights of cannon shells.

  Desch’s lab had become a technology showplace for NCR, drawing visitors from all over the world. Edward A. Deeds and Charles F. Kettering often stopped by to check on their pet projects. So did the dean of the Dayton engineering pantheon, Orville Wright. Wright and Desch were climbing the steps to Building 10 one day in 1940 when a bomber flew overhead on its way to landing at Wright Field, the local base that bore the name of Orville and his brother Wilbur. Hobbled by a limp suffered in an early flying accident, Wright had stopped on the steps and was stari
ng up at the plane when Desch worked up the nerve to ask the aviation pioneer if it troubled him that his invention had been used for war.

  Wright reflected a moment, then shook his head. “No,” he said. “If we hadn’t done it, somebody else would have. It was inevitable.”

  Desch ran into his own inevitabilities in the summer of 1942. The more he went over the Navy plans, the more he bumped up against an inescapable conclusion: an all-electronic machine was impossible. If the electronic American Bombe followed the architecture of Britain’s, it would devour more than twenty thousand tubes run at speeds that couldn’t be sustained. And a universal codebreaking machine, one that could attack more than just the Enigma, would need thousands more tubes and even higher speeds. Either machine would demand too many scarce resources to manufacture, would generate too much heat to operate reliably, and would utilize more electrical power than could be supplied by conventional means.

  Wenger, Engstrom, and Howard’s MIT group were devastated, but Desch was highly respected, and his judgment could not be ignored. Yet it meant the entire RAM project was in danger, as was perhaps even OP20G’s research section.

  OP20G’s promised electronic Bombe was now a failure, its second RAM project was stalled, and it was contributing little to the war effort in the Atlantic. There was only one solution: turn the whole problem over to Desch. He would have to be drawn into the tight circle of people who knew of America’s and Britain’s great secret, Ultra.

  How Desch felt about being privy to one of the most important Allied strategic advantages of the war is best gauged by his reaction more than thirty years later, when he learned the British were divulging their role in Ultra with the publication of F. W. Winterbotham’s book The Ultra Secret. Desch’s daughter, Debbie Anderson, recalled that her father launched into a stream of invective against the British, yelling, “I always knew those SOBs would talk!” He took his own part in that secret to his grave.

  Later that summer, Desch began receiving new details about the design of the British Bombe, supplied by Robert Ely and Joe Eachus, the U.S. Navy officers who had been installed as the first American codebreakers at Bletchley Park. The two began sending coded telegrams back to OP20G almost as soon as they arrived at Bletchley in June, and the information kept flowing from England throughout the summer.

  Eachus, who was thirty-two at the time, was an Indiana native raised in southwest Ohio. After graduating from Miami University, near Dayton, and earning his Ph.D. in mathematics from the University of Illinois, he had been teaching at Purdue and, almost as a lark, taking the Navy’s correspondence course in cryptanalysis when the Japanese attacked Pearl Harbor. Soon after, Eachus went into the Navy. A little more than a year later, he was one of two OP20G officers assigned to work temporarily at Bletchley, side by side with Turing and the other British codebreakers. Eachus had no complaints about British cooperativeness: “As a matter of fact, I remember some of the places I visited, the people were anxious to talk about what they were doing because they had almost no opportunity to do so.”

  One of the first vital pieces of information to come back to the United States was the need for a two-way flow of electricity through the Bombe’s commutator wheels, a Turing invention that gave the Bombe much of its problem-solving power. Theoreticians at OP20G believed that two-way electronic tubes could perform the same task, at blinding speeds, but Desch was later to inform them otherwise.

  The evidence indicates that the British told Ely and Eachus everything about their existing Bombes, including the cold-point tests that greatly increased their efficiency, as well as what was known at GCCS at the time about the forthcoming four-wheel Bombes. Blueprints for the four-wheel design, as it existed in the summer of 1942, were sent to Washington sometime in September. How much and exactly what Desch was told by OP20G during the summer remains unclear. But on their return from England in August, Ely and Eachus reported to Desch’s supervisors in Washington, including Howard Engstrom, and later made trips to Dayton themselves.

  Desch and Engstrom had agreed that the Dayton Bombes would follow the general logic and technology of Britain’s machines, but how many Bombes were to be built, how they were to be operated, and how much speed and codebreaking capability was to be designed into each machine were topics still under debate. One option was to build 336 of the two-and-a-half-ton machines, to run through all possible Enigma-rotor combinations without changing wheels, thus saving operating time and labor but greatly increasing the cost of the project. Other questions concerned how many Bombes should fit into a single frame and how many Enigma machines to simulate in each Bombe. A greater number of Enigma analogs per Bombe would give them more codebreaking power but also would demand more electrical and mechanical power to drive the additional wheels at the necessary speeds.

  A more general question was how many automatic features should be built into the machines. More features would help meet the demands of those who would run the machines and produce intelligence. But fewer would lower the cost of construction and improve reliability and performance.

  Each answer would engender a different set of engineering challenges for Desch and his team.

  BY MID-SEPTEMBER, Desch was able to go back to his Navy supervisors with a plan for an original American Bombe that could approach the performance, if not the elegance, of the Navy’s all-electronic design. Desch’s machine, part electronic but mostly mechanical, would attempt to crack Shark by fusing Britain’s Bombe logic with the best of America’s high-speed electromechanics. Still, it would have to make use of the labor-intensive British-style commutators, which simulated the Enigma code wheels and had to be changed with each run of the machine. But such a Bombe could be operational by the end of the year, and, more important, production could begin in a matter of months.

  Desch’s memo of September 15, 1942, outlining his machine in less than twelve pages, is a marvel of clarity and conciseness. Yet it reveals some startling gaps in his knowledge about the workings of the British Bombe and its supporting methods, leading him to wildly overestimate the number of stops or false hits (three thousand for each test of the Enigma’s wheel order) and hence to overstate the number of Bombes that would be required. *21

  No doubt at the request of British codebreakers, Desch incorporated into the design the ability to run three-wheel as well as four-wheel Enigma problems, in case the British should need help against the German Army and Air Force ciphers. To speed the Bombe’s operation, he included an automatic feature that would record the locations where the fast-spinning wheels had found a possible hit and, after stopping them, rewind them back to their hit positions for conducting the additional Diagonal Board test. That way, a new search for possible Enigma solutions could begin immediately with no gaps in coverage. “A small motor, in addition to the main drive motor, could perform this function,” he wrote. “A rewind speed of about 100 rpm probably could be realized. The bomb then would be completely automatic.”

  But the innovative heart of the machine would be its electronic tracking system—a truly original Desch contribution, which he described in just one sentence: “The memory device will consist of two banks of 26 miniature thyratrons each, one tube in each bank being ignited when a hit occurs.”

  Desch’s report estimated that the Navy would need 336 of the costly machines—nearly three times more than later experience was to indicate. The huge error, corrected in the following months, was based on Desch’s still incomplete knowledge of the British Bombe. Even so, the Navy brass gave him the go-ahead for the inflated figure in his report, perhaps because, with that many machines, there would be no need for time-consuming wheel changes between runs.

  OP20G was desperate and willing to take a huge gamble on the design. The Navy promised everything and anything Desch needed to get the job done as quickly as possible—millions in funding, hundreds of trained personnel, and the highest top secret priority. The old engineering group from MIT was ordered to put the other machine projects on hold until the
critical Bombes were ready. Immediately, Desch hired ten more top civilian engineers and put them to work on a trial machine.

  In early October, Wenger was so confident of the project’s success that he promised the first Bombes by the beginning of 1943. He began negotiations for the land and buildings to house the Bombes once they came off the production line, as well as the hundreds of clerks, maintenance men, and armed guards who would be needed for their operation.

  WITH ULTRA AND the Battle of the Atlantic in the balance in 1943, the Bombe project was placed on a fast track, and cost overruns in the millions were accepted without question. The jump from an estimate of two million dollars to one of four million in a few months did not threaten the program. Resources were seldom a problem at NCR, despite the heavy demands from other war industries across the country. At its height, the project employed more than one thousand manufacturing workers and required material and components from thousands of different suppliers. And the greatest efforts were made to keep the project secret and secure, given the fears that even a hint of the project’s aims would cause the Germans to alter their code systems. On March 17, 1943, when the German U-boat threat was heightened by another Enigma blackout, the Navy asked Roosevelt for the highest possible priority for the project, the president’s AAA designation. He granted it the very next day.

  At times, OP20G had to bow to the personnel demands of other branches of the Navy in order to guard the project’s security: relying too often on its AAA status might tip off the wrong people to the American Bombe and the Ultra secret. OP20G thus found it almost impossible to persuade the Navy that sailors should be transferred from sea duty to technical detail on an inland project.