The Community for Technology Leaders

Events and Sightings

Pages: 92–96

Abstract—This Events & Sightings installment covers a range of recent events focusing on the history of computing.

Keywords—history of computing; Project MAC; Multics operating system; Norbert Wiener; Differential Analyzer; Tokyo University of Science

Restoration of the Differential Analyzer at the Tokyo University of Science

At one time, there were at least three mechanical differential analyzers in Japan, installed at the University of Tokyo (4 integrator units, 1942?), University of Osaka (3integrator units, 1944?), and University of Tokyo (8integrator units, 1954). Later, the Osaka machine was sent to the Tokyo University of Science and kept in its Museum of Science for exhibition. It was certified as an Information Processing Technology Heritage in 2009. The two other analyzers were abandoned.

Hearing that the Manchester Museum of Science and Technology restored its differential analyzer, we were tempted to resurrect our survivor as well. After determining that a “heritage” computer could actually be modified, in April 2013, a new project was kicked off to restore the differential analyzer to working condition. The project core was made up of the Tokyo University of Science, National Institute of Informatics (NII), and National Institute of Information and Communications Technology (NICT).

The first step was to examine the current state of the machine. Since the machine was assembled after being moved from Osaka by nonengineering people, there were many components incorrectly connected. To keep parts from rusting, shafts, gears, and adders had been painted with layers of lacquer. In addition, the driving belts of the torque amplifiers were broken, and disc surfaces were scratched. It was quite a miserable sight. On 15 June, a team of students from the Mechanical Engineering Department was summoned to the museum to measure every dimension of the machine and to draw precise plans. After that, two of the three sets of integrator units and torque amplifiers were removed from the bay frames and taken to the NICT machine shop for repair.

On 20 September, the remaining shafts and gears were removed from the bay frames, dipped in a vat filled with thinner, and kept there for two or three days and nights. The student team was then asked to remove the lacquer by patiently washing each part. The cleaned shafts and gears were returned to the original positions in the bay frames. Then they were turned freely. In August, at the NICT machine shop, the first integrator unit was thoroughly studied and all parts were removed, cleaned, and checked to see if they were worn out. Some parts were missing. In those cases, the worn or lost parts had to be made anew.

For the large integrator disc to be reformed, its axis was clamped on a turning lathe, and the surface was scraped carefully until a flat plain appeared. Parts of the torque amplifier were also completely removed. The drum surfaces of the primary and secondary stages were ground, and in so doing, a few prototypes of string holder mechanisms were designed and tested. Because the designers foresaw that the strings would need to be adjusted often, and the space in the torque amplifier was so small, the string holder could be handled easily. However, the material of the friction strings posed a problem. In an interview, Professor Emeritus Masaru Watanabe, who had actual experience with both of the differential analyzers at the University of Tokyo, recalled that they used koto string for it long ago. (Koto is a Japanese classical musical string instrument.) Koto string is still available, so it was chosen for the first stage. For the secondary stage drum, kevlar cord, a kind of poly chemical substance, was the sole solution. The first torque amplifier and integrator set was combined in December, although the output torque proved far weaker than expected.

In January and February 2014, the amplification ratio was measured by repeatedly changing the tension, friction, and number of turns. Wrapped with sample string, the drum was rotated slowly on a lathe, and both ends of the string were pulled by a pair of spring scales to calculate the ratio. Finally, an amplification on the order of a thousand was achieved.

By the beginning of April 2014, the first set was complete, so the restoration for the second set of integrator and amplifier began. This process moved rapidly, and the second set was completed in a few days. The two sets of integrator units and torque amplifiers were transferred back to the Museum of Science on 9 April. Laborious adjustments between the gears of integrator and cross shafts were needed, by raising, lowering, or tilting little by little, before the first set became operational. The typical test with a single integrator was an exponential function, so shafts and gears were set up for $_$\int {y}\,dx = y$_$. The torque amplifier motor was switched on, followed by the independent variable motor, and the output value seemed to increase exponentially. Bravo! A fortnight later, the second set became ready for operation. Shafts and gears were set up for the circle test, an example well known by differential analyzer geeks. All motors were turned on. The restored differential analyzer worked beautifully. For a complete circle to be drawn, it took about 6 minutes and 40 seconds (see Figure 1).

Graphic: Tokyo University of Science differential analyzer. Following an extensive restoration project, it was returned to working order.

Figure 1. Tokyo University of Science differential analyzer. Following an extensive restoration project, it was returned to working order.

Since the output table had not yet been repaired, two rotary encoders were connected to the corresponding cross shafts. The output of the encoders was picked up by a PC, and a graph similar to that of the output table was drawn by the PC. Shortly after the whole machine became workable, the Museum of Science held an exhibition of analog computers. Needless to say, the differential analyzer was one of the highlights. Dizzy movements of discs, wheels, shafts, and gears gave visitors a strong impression of the differential equations being solved by means of mechanical integration (see Figure 2).

Graphic: Restored differential analyzer. The working system was exhibited to members of the Board of Directors from the Information Processing Society of Japan.

Figure 2. Restored differential analyzer. The working system was exhibited to members of the Board of Directors from the Information Processing Society of Japan.

The members who undertook the restoration understood well the principles behind the differential analyzer. Nevertheless, the real machine proved to be far more complicated, with various sorts of tricky gadgets to insure stability and accuracy such as a lashlock, flywheel, and frontlash unit. We discovered many engineering techniques from the past two or three generations.

In conclusion, it must be stressed that thisrestoration project was entirely supported by the NICT machine shop. The machine shop possesses a fleet of machine tools, preserves every kind of raw material, and stocks many machine parts. Each time new pieces of any part were requested, they were supplied instantly, à la Aladdin's lamp. This reminded me of Arthur Porter's recollection: “Fortunately, the physics laboratories were equipped with excellent machine-shop capabilities, and I had no difficulty in building the apparatus and performing the investigations.”1 Our project was similarly fortunate because of the cooperation of the NICT machine shop.


Eiiti Wada is a research advisor at the Research Laboratory of the Internet Initiative Japan and a professor emeritus at the University of Tokyo. Contact him at
Tom Van Vleck is an independent consultant, working in Web applications, security, cloud computing, data mining, and programming. From 1965 to 1974, he was a manager and system programmer at the Massachusetts Institute of Technology, where he participated in the Project MAC development. Contact him at
Rudolf Seising is acting as a professor for the history of science at the Friedrich-Schiller-University of Jena, Germany, and he is an affiliated researcher at the European Center for Soft Computing (ECSC) in Mieres, Spain. Contact him at
65 ms
(Ver 3.x)