Issue No. 04 - October-December (2005 vol. 27)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MAHC.2005.53
Laurie Robertson , Editor, Virginia Tech
Reflections on the Difference Engine
Doron Swade's The Difference Engine1 describes Charles Babbage's early thoughts on table creation and the use of finite differences, the origins of the Analytical Engine, and the 1991 reconstruction of the Difference Engine at the Science Museum in London in time for Babbage's 200th birthday anniversary.
Some of Swade's new revelations are worthy of comment. First is the problem that Babbage encountered in designing a mechanical system for simultaneous carry propagation in which he had to eliminate both the ripple delay resulting from sequential propagation of carries and the large torque required to add 1 to 999999999, for example. Babbage produced several designs to overcome these problems, but interestingly, when the Swade group attempted to build the device, it found that some of the drawings were mirror images of the correct components. Swade discusses possible causes of this error but gives no reasonable solution. One possibility might be that the drawings were intended for reproduction by the old office letterpress method so that copies could be given to the workshop and the copy would correct the inversion.
The second revelation was the complex physical arrangements Babbage needed to get enough space between the number storage stacks and the adding mechanisms so that adjustments and repairs could be made. This led Swade to comment on Babbage's apparent dismissal of number systems other than decimal. Swade then mentions a possible ternary mechanism proposed by an English mechanic, Thomas Fowler, from 1840 to 1842. 2 A simple model and calculation shows that the optimum number base for minimizing complexity is e. This is of course impossible and the nearest practicable bases are 2 and 3.
My Automatic Relay Computer
All these things brought to mind my own earliest digital computer, the Automatic Relay Computer (ARC). I designed this between May and July 1947 while working with the von Neumann group at the Institute for Advanced Study in Princeton. It was a fully parallel, 21-bit, relay machine; the only electronic vacuum tubes were in the magnetic drum store. The architecture was von Neumann in the sense that it consisted of input–output, control, and an arithmetic unit. It differed in that it used 21-bit words and two-address coding to enable effective use of the drum store by minimizing access time. My diary notes show that I anticipated the same delay problem with carry propagation that Babbage faced, and I produced a relay version of the adder that overcame it.
Design with relays was simple and certain to work, but the magnetic storage unit had yet to be developed. In the meantime, I asked my father, Sidney J. Booth, a mechanical engineer, to design a mechanical equivalent to the magnetic store using metal pins that could be moved to two positions by simple electromagnets; following the design, he had this device constructed. Digits were stored in a channel along the surface of a 2-inch diameter brass cylinder. Each digit was separated from its neighbor by a grooved recess. These pins could be moved into two positions by levers operated by small electromagnets (see Figure 1). In the set position, the pin operated a contact to generate the appropriate digit. In the unset condition, the contact was left open. The device could be operated continuously or sent to a desired position using the "clock" track seen on the right-hand end of the drum.
On my return to England in September 1947, the relay part of the machine was constructed by my two assistants Kathleen Britten (a professor at Lakehead University, Ontario, and Xenia Sweeting). It worked perfectly and, with a prototype magnetic drum (see Figure 2), I demonstrated it to members of the Board of the British Rubber Producers' Association on 12 May 1947. The demonstration elicited continued support for my future machines.
A search of my old records revealed a moldy file, which contained some of my design sketches and notes from Princeton in the 1940s. There are sketches of the original relay adder and the version with simultaneous carry. The latter consists of a single relay (see Figure 2).
The relays have a 1-millisecond contact transit time (normal relays require 10 to 20 ms). They have two coils, one on each side of a U-shaped yoke. The coils are connected in a series, but it is simple to attach a lead to their juncture. The two activating points are labeled A and B. The center lead is grounded. Voltage applied to A or B separately will turn the relay on, both together will produce the off condition. The relay is thus an exclusive OR device. Input digits are applied to coils A and B. Carry from a previous stage is connected to the relay's moveable contact. Input on A or B (but not both) means that an input carry generates an output carry, inputs on A and B leave the relay off. The off contact point is connected to B and thus initiates a carry directly if B is live. Combining these with a second relay generates the sum digit. Figure 3 shows the carry circuit.
The actual unit consists of 10 relays to enable storage of the various digits and left and right shift operations.
ARC's original 21-digit plus clock magnetic drum 3 is in the Science Museum in London. It is not quite in its original form because it was cannibalized for another student project in Fourier synthesis. 4Figure 4 shows the original ARC drum.
The mechanical pin drum is in the Museum of Technology in Birmingham.
Unfortunately, I was too busy to write up the ARC and, although I left it at Birkbeck College, London, when I moved to the US, all trace of it is now lost. Figure 5 shows the control and arithmetic units just after they were operational in 1948.
This nostalgic trip has caused me to wonder, as I have in the past, why Babbage did not try to design and construct a relay version of his Analytical Engine. Samuel Morse, of Morse Code fame, had already constructed a relay and described its use in data switching in 1836. 3 After all, iron and copper wire were easily available, and there was no need for great speed. Fowler's storage idea could have been extended, easily, to store binary digits electrically.
Andrew D. Booth
The founders of the ACM
The recent death of Richard van Duyne Campbell, one of the eight founders of the Association for Computing Machinery (ACM), reminded me that the available information about all the founders is thin, scattered, and incomplete. Indeed, for one of the group, R. Taylor, ACM records show only his last name and first initial. I have assembled some annotated basic information about the founders, all of whom I knew, as a tribute to them.
Figure 6 lists the members of the self-selected temporary committee, which Franz L. Alt 1 credits with founding the ACM by making a call for an association in a document dated 25 June 1947. 2
The founders were all about the same age; John W. Mauchly was the oldest at 40, and Richard V.D. Campbell was the youngest at 31. The war had interrupted their barely started professional lives, but computing machinery now offered them new and better careers. These were "Young Turks," dissatisfied with what their elders were doing, setting out to support their new discipline with its own organization. Years later, spark-plug Edmund Callis Berkeley told me that he knew that he needed at least one big name on the committee and made several approaches. Harvard associate professor, Howard H. Aiken, of the Harvard Mark I calculator and Berkeley's former boss, refused to sign on, saying that computing was a subbranch of mathematics, which had enough organizations. Berkeley never forgot this rebuff, which he took as a personal insult and added it to the list of grievances he held against Aiken. University of Pennsylvania professor John Grist Brainerd, of ENIAC and the Moore School, also declined saying that computing was a subbranch of electrical engineering, which at that time had both the American Institute of Electrical Engineers and the Institute of Radio Engineers. John W. Mauchly, of ENIAC, Berkeley's third choice, was enthusiastic about his plan and later served as an ACM officer for several years. Mauchly's ENIAC partner, J. Presper Eckert, declined to be a founder on the same grounds as Brainerd. These invitations by Berkeley as well as ACM itself are unmentioned in later accounts of the lives of the men who declined to sign his call for action.
All the founders except John H. Curtiss had wartime experience with computing machinery. Berkeley, Campbell, and Harry E. Goheen were all members of Aiken's Harvard Mark I team. Mauchly and T. Kite Sharpless participated in designing and building the ENIAC. Charles B. Tompkins had used the secret Navy code-breaking machines. Richard Taylor, assistant professor at the Massachusetts Institute of Technology, had worked on and used both of the Bush differential analyzers. Although Curtiss had no hands-on computing experience, he was full of enthusiasm for the application of computing machinery to mathematical and statistical problems.
The committee members represented three traditional scientific or technical disciplines: Mauchly was a classical physicist; Berkeley, Curtiss, Tompkins, and Goheen were mathematicians in the broad sense of the term; and Campbell, Sharpless, and Taylor were electrical engineers. This curious combination of mathematicians and electrical engineers has always been present in ACM.
The military connection
The charge some historians have made that computing was actually created by and for the military can be applied to ACM itself for five of the founders; Berkeley, Campbell, Goheen, Tompkins, and Curtiss had been naval reserve officers on active duty during the war. None of them, however, had been salty warriors in the sense of going to sea and shooting guns, and when the war was over, they all reverted promptly to the rank of civilian.
All the firms and the two universities were leaders in computing at the time. Their geographical grouping is reflected in the name the founders selected, The Eastern Association for Computing Machinery, a name which only lasted until January 1948.
All the founders continued in computing. Berkeley became an author, editor, and publisher of popular computing material. Campbell was a lifelong computer engineer. Curtiss developed the National Bureau of Standards into a leader in computing. Goheen and Tompkins were professors of mathematics and computing. Mauchly was a leader in computing on his own and with Univac at Remington Rand. Sharpless formed his own successful computer firm, Technitrol. Taylor earned a PhD and then joined IBM at Endicott, where his trail disappears because IBM put all its personnel records on computers and destroyed the paper records of employees who had retired before 1980.
Except for Taylor, all the founders took part in ACM's development, and 10 years later, in 1957, all except Taylor were still members. In 1972, all the living founders were invited to a gala cocktail reception and full-course banquet in honor of ACM's silver anniversary. Figure 7 shows how they were listed in the banquet's program.
Only Berkeley, Goheen, and Campbell were ACM members. Berkeley Enterprises published computer books and periodicals, Dynatrend was a computer-consulting firm, and Mitre Corp. was part of the industrial-military complex with a computing specialty.
Berkeley, who the ACM formally identified with some pomp and ceremony at that time as its singular founder, made the keynote speech at the banquet. He said that it was a "gross neglect of responsibility" that ACM did not have committees investigating whether computer applications were good or evil. He encouraged data-processing professionals to use "social enterprise" to head off his prediction that mankind would be extinct in 500 years. He said that use of nuclear weapons and irreversible environmental changes, such as an increasing amount of carbon dioxide in the atmosphere, made the situation "too hard to analyze." His ferocity increased as he predicted that vested interests of large corporations would "checkmate" any possible solutions, and he called for the formation of an "Association for the Prevention of Doomsday." He said that the use of computers in the Vietnam War made him "ashamed of belonging to the computer field."
His audience became increasingly restive as his condemnations became specific, and when he finally criticized Honeywell, by name, for its "atrocity engineering" in designing antipersonnel bombs, several prominent members followed Grace Murray Hopper as she ostentatiously stood up and walked out while he was speaking. That left me, as toastmaster, to tell a diminished but bestirred audience to charge their glasses and respond to my toast, "Our Founder."
Although Computerworld summarized Berkeley's speech, 3 the ACM did not publish it and never referred to it in any way at that or any later date—an action reminiscent of Orwell's "Memory Hole."
By 1984, only Berkeley and Campbell were still ACM members, although Goheen and Tompkins were still alive. I believe that Berkeley and Campbell continued to be members until they died, although Campbell, who lived into the era of ACM Fellows, was never so honored.
Eight real men who should not be forgotten founded ACM. I would appreciate hearing from anyone who can tell me more about any of them.
References and notes
Eric A. Weiss