Milestone-Proposal:Parametron, 1954
To see comments, or add a comment to this discussion, click here.
Docket #:2025-07
This proposal has been submitted for review.
To the proposer’s knowledge, is this achievement subject to litigation? No
Is the achievement you are proposing more than 25 years old? Yes
Is the achievement you are proposing within IEEE’s designated fields as defined by IEEE Bylaw I-104.11, namely: Engineering, Computer Sciences and Information Technology, Physical Sciences, Biological and Medical Sciences, Mathematics, Technical Communications, Education, Management, and Law and Policy. Yes
Did the achievement provide a meaningful benefit for humanity? Yes
Was it of at least regional importance? Yes
Has an IEEE Organizational Unit agreed to pay for the milestone plaque(s)? Yes
Has the IEEE Section(s) in which the plaque(s) will be located agreed to arrange the dedication ceremony? Yes
Has the IEEE Section in which the milestone is located agreed to take responsibility for the plaque after it is dedicated? Yes
Has the owner of the site agreed to have it designated as an IEEE Milestone? Yes
Year or range of years in which the achievement occurred:
1954
Title of the proposed milestone:
Parametron, 1954
Plaque citation summarizing the achievement and its significance; if personal name(s) are included, such name(s) must follow the achievement itself in the citation wording: Text absolutely limited by plaque dimensions to 70 words; 60 is preferable for aesthetic reasons.
The Parametron, a logic element using ferrite cores and parametric oscillation, was invented in 1954 by Eiichi Goto at the University of Tokyo. 4,200 parametrons went into the PC-1, Japan’s first university-built stored-program computer which became the nation's then-fastest in 1958. The parametron’s low cost and electrical stability shaped Japan’s early computer development and scientific research, and nurtured the country's first generation of computer engineers.
200-250 word abstract describing the significance of the technical achievement being proposed, the person(s) involved, historical context, humanitarian and social impact, as well as any possible controversies the advocate might need to review.
Parametron, invented in 1954 by Eiichi Goto, a graduate student at the University of Tokyo, is a logic element leveraging ferrite cores and parametric excitation for reliable, low-cost logical operations. By encoding binary states (0/1) through the phase of oscillation in a resonant circuit, it provided a stable alternative to the expensive vacuum tubes and unreliable transistors of 1950s computing. This innovation enabled 'majority rule logic,' supporting diverse computational functions with remarkable efficiency.
Separately, Goto’s work on Parametron and his other innovations contributed to the PC-1, Japan’s first stored-program computer at a university, completed in 1958. Built with approximately 4,200 parametrons, the PC-1 was the nation's fastest computer of that time, featuring an arithmetic unit with high-speed carry propagation, concurrent execution of two instructions for enhanced speed, a pioneering AC-driven magnetic core memory with non-destructive reading and error correction for access, alongside interrupt-driven multitasking. This breakthrough democratized electronic computing for natural science researchers across Japanese universities, who depended on it as a free resource.
Though overtaken by transistor advancements in the 1960s, the parametron’s influence persisted. It jumpstarted Japan’s computing infrastructure and fueled early scientific research, securing its place in technological history. This invention highlights how creative design can overcome practical limitations, leaving a lasting legacy in computing evolution.
IEEE technical societies and technical councils within whose fields of interest the Milestone proposal resides.
IEEE Computer Society
In what IEEE section(s) does it reside?
IEEE Tokyo Section
IEEE Organizational Unit(s) which have agreed to sponsor the Milestone:
IEEE Organizational Unit(s) paying for milestone plaque(s):
Unit: IEEE Tokyo Section
Senior Officer Name: Toshiro Hiramoto
IEEE Organizational Unit(s) arranging the dedication ceremony:
Unit: IEEE Tokyo Section
Senior Officer Name: Toshiro Hiramoto
IEEE section(s) monitoring the plaque(s):
IEEE Section: IEEE Tokyo Section
IEEE Section Chair name: Toshiro Hiramoto
Milestone proposer(s):
Proposer name: Chiaki Ishikawa
Proposer email: Proposer's email masked to public
Please note: your email address and contact information will be masked on the website for privacy reasons. Only IEEE History Center Staff will be able to view the email address.
Street address(es) and GPS coordinates in decimal form of the intended milestone plaque site(s):
Science Gallery, 1st floor (ground floor) of Building 1, Faculty of Science, The University of Tokyo. Address; 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033 Japan.
GPS coordinates; 35.7137465,139.7608399
Describe briefly the intended site(s) of the milestone plaque(s). The intended site(s) must have a direct connection with the achievement (e.g. where developed, invented, tested, demonstrated, installed, or operated, etc.). A museum where a device or example of the technology is displayed, or the university where the inventor studied, are not, in themselves, sufficient connection for a milestone plaque.
Please give the address(es) of the plaque site(s) (GPS coordinates if you have them). Also please give the details of the mounting, i.e. on the outside of the building, in the ground floor entrance hall, on a plinth on the grounds, etc. If visitors to the plaque site will need to go through security, or make an appointment, please give the contact information visitors will need. Building 1 of Faculty of Science, the University of Tokyo is where parametron was invented. There is now a space to display many discoveries made by the researchers there and this Milestone will be displayed there.
Are the original buildings extant?
No. The building was rebuilt about a few dozen years ago.
Details of the plaque mounting:
It is placed on the wall of so-called "Science Gallery" where many displays related to science are placed. See: https://www.s.u-tokyo.ac.jp/en/gallery/
How is the site protected/secured, and in what ways is it accessible to the public?
Visitors can come to the Science Gallery without security check.
Who is the present owner of the site(s)?
Shin-ichi Ohkoshi Dean of School of Science, the University of Tokyo
What is the historical significance of the work (its technological, scientific, or social importance)? If personal names are included in citation, include detailed support at the end of this section preceded by "Justification for Inclusion of Name(s)". (see section 6 of Milestone Guidelines)
Justification of name-in-citation
Eiichi Goto: Pioneer of Parametron Technology and Early Computers
Eiichi Goto (1931–2005), a Japanese scientist, invented the parametron in 1954 as a graduate student and spearheaded the development of the PC-1, the first fully programmable stored-program computer at a Japanese university to utilize this technology, completed in 1958. His innovations catalyzed advancements in Japan’s post-war computer industry, enhancing scientific research and industrial capabilities during a critical period of technological recovery.
Invention of the Parametron
In 1954, Eiichi Goto invented the parametron, a logic device leveraging nonlinear parametric oscillation with two ferrite cores [9, 16, P1–P2]. Unlike the vacuum tube and early transistor circuits prevalent at the time, the parametron offered remarkable stability, requiring minimal maintenance compared to vacuum tubes with short lifetime and costing significantly less than both vacuum tubes and nascent transistors. Its simplicity and reliability made it an ideal foundation for computer design. Early applications showcased its superior fault tolerance over competing technologies, such as vacuum tubes with relatively short-lifetime, slow electromechanical relays, and unstable point-contact transistors. Driven by a desire to construct an electronic computer within the constrained budget of Takahasi’s laboratory at the University of Tokyo, Goto and his mentor, Takahasi, exhaustively had explored affordable solutions; the parametron emerged as a breakthrough, laying a critical foundation for Japan’s post-war computing advancements.
Development of the PC-1 Parametron Computer
In 1957, Eiichi Goto led the development of the PC-1, a pioneering fully programmable stored-program computer powered by 4,200 parametrons [4–9, 11]. PC-1 began its operation on March 26, 1958.
Representing a significant advancement in post-war Japanese computing, the PC-1 outperformed emerging transistor-based systems in stability and featured a meticulously crafted instruction set, created by study of machines like the EDSAC ([12] The Preparation of Programs for an Electronic Digital Computer, M. V. Wilkes et al., 1951) by Goto, Takahasi and the rest of the team. The parametron’s inherent stability and multi-input capability enabled reliable, straightforward logic circuit design, exemplified by the fast arithmetic circuit that utilizes carry select mechanism, which can handle the carry propagation in O(sqrt (N)) logic step or time where N is the bit length, that took advantage of majority logic, and the 1959 implementation of an interrupt function—among the earliest of its kind—facilitating multitasking between the main program and I/O handling. Unlike the complex, maintenance-intensive vacuum tube computers of the era (vacuum tubes had relatively short lifetime compared with other circuit element, thus the necessity of replacement of a few vacuum tubes a day for constant operation), the PC-1 achieved high reliability with fewer components, low power consumption, and exceptional fault tolerance, ensuring sustained operation in a university laboratory environment.
Photo 1 Eiichi Goto and the PC-1 Parametron Computer (Source: Information Processing Society of Japan)
Photo 2 Eiichi Goto adjusting the memory section of Parametron
computer PC-1 (Source: Information Processing Society of Japan)
PC-1 was used by many scientific researchers after it became
operational in March 1958. It had profound impact on the Japanese
computer industry and scientific community. Details of the PC-1,
especially the software side, are thoroughly described in Eiichi
Wada's lecture materials from the Parametron Memorial Lecture held in
2008, commemorating 50 years since PC-1's implementation [11].
Libraries and calculation done on PC-1 are summarized in Appendix II.
Influence of Parametron Technology and Goto's Achievements
The parametron’s affordability, minimal maintenance, and operational stability swiftly captured Japanese industry's attention following Goto’s initial presentation at a Japanese conference. Commercial entities soon adopted the technology, developing and marketing computers and calculators based on it (see Appendix I for examples). Some models sold more than 700 units back then.
At a time when vacuum tubes with the inherent short life cycle demanded frequent upkeep and early transistors proved costly and unreliable, the parametron filled a critical niche, earning Goto the prestigious IRE Browder J. Thompson Memorial Prize in 1961 [A1] for his seminal paper, “The Parametron, a Digital Computing Element Which Utilizes Parametric Oscillation” [Proc. IRE, Aug. 1959]. This award underscored the technology’s profound impression on both industry and academia, cementing Goto’s reputation as a visionary innovator.
In Japan, he was awarded the prestigious Asahi Prize in 1959. This was for his contribution to PC-1. [A2] The Asahi Prize (朝日賞, Asahi Shō), established in 1929, is a prize presented by the Japanese newspaper Asahi Shimbun and Asahi Shimbun Foundation to honor individuals and groups that have made outstanding accomplishments in the fields of arts and academics and have greatly contributed to the development and progress of Japanese culture and society at large. (from Wikipedia, https://en.wikipedia.org/wiki/Asahi_Prize).
Conclusion
The parametron profoundly shaped Japan’s post-war computer industry, with its influence extending internationally through U.S. patents [US Patent 2,948,818 and 2,948,819]. As its sole inventor, Eiichi Goto elucidated its principles in his landmark IRE paper and led the Takahasi Laboratory team at the University of Tokyo in creating the PC-1. Given his pivotal role in this enduring technological achievement, Goto’s name merits inclusion in the IEEE Milestone citation, honoring his lasting legacy in computing history.
Historical Significance
Background
The parametron is a logic element that utilizes the parametric excitation phenomenon, leveraging the non-linear magnetic response of ferrite cores. It was invented in 1954 by Eiichi Goto, who was then a graduate student at the University of Tokyo's Graduate School of Science.
In the 1950s, when computers were in their infancy, constructing a flip-flop in Japan cost around 1,000 yen for a vacuum tube and several thousand yen for a transistor. At the time, point-contact transistors were unreliable and unstable. The earliest junction-type transistors coexisted with point-contact transistors and were slow, so despite the instability, point-contact types were used for computers.
At the Takahasi Laboratory of the University of Tokyo, where Goto
studied, there was a keen interest in computational machines. Various
devices, such as a computer using a rotary switch from a telephone
exchange and a decimal computer using decatron tubes, were examined
and manually simulated.
(An earlier study of information storage devices [1], and [2] are
examined in Appendix IV to give an overview of the kind of
theoretical and experimental study that were done before the invention
of Parametron.)
Eiichi Goto's knowledge of physics and applied
mathematics was invaluable during this time. Consideration was given
to using ferrite cores [3], which cost only 5 yen each, leading to the
concept of utilizing the parametric excitation phenomenon.
Consequently, the element was named the parametron.
Inexpensive ferrite core
Back then (pre 1954), a vacuum tube cost 1,000 Japanese yen, and a transistor 8,000 yen. A ferrite core was mere 5 yen. Goto mentioned the price situation in a magazine interview (in Japanese) quoted below. The following English translation by the submitter is an excerpt from the interview in Japanese.: https://ascii.jp/elem/000/001/221/1221954/2/
——Why did you create parametron? At the time, there were no computers in Japan.*1 So we decided to make one at the Takahasi Laboratory at the University of Tokyo. But the university's research budget was not enough. At the time, vacuum tubes cost 1,000 yen each, and transistors cost 8,000 yen each. Also, vacuum tubes wear out quickly. Transistors were still unstable. In comparison, ferrite cores, which had long been familiar in, say radio circuit construction, cost just 5 yen each. So I thought, why not use ferrite cores as the material? Ferrite cores are a stable material, and since they are made of pottery, they don't break if you make them right the first time. The name “parametron” was given because it uses the principle of parametric excitation.
*1 The first Japanese vacuum tube stored program computer was built in 1956 (FUJIC). https://museum.ipsj.or.jp/en/computer/dawn/0010.html More on FUJIC in Appendix V for speed comparison.
The first Japanese transistor stored program computer was built also in 1956 (ETL Mark III): https://museum.ipsj.or.jp/en/computer/dawn/0011.html
UNIVAC 120 was the first commercial electronic computer installed in Japan in 1955.: https://museum.ipsj.or.jp/heritage/UNIVAC120.html
So, as Goto mentioned "There were no electronic computers in Japan" before 1954. Note: The submitter thinks Goto's recollection of 1,000 yen is a ballpark figure. But the two order of magnitude difference (factor of 200) is enough to show the cost merit of using ferrite cores.
A typical single parametric logic element used three cores (one larger
core was added to merge the signals from input wires from multiple sources) , one
capacitor and a resistor. Three such units were combined to create a
circuit element to pass logic signal to the next parametron unit.
Such a combination would NOT reach the price of a single vacuum tube
alone. And the vacuum tube circuit also requires capacitors and
resistors to boot. The difference of 1,000 Yen and 5 yen of main component is
large.
Takahasi recalls that parametron element after assembly was sold at 500 YEN. He stated that a similar logic element device made of vacuum tubes would have costed more than 10,000 YEN, and the similar unit made of transistors would have costed too much to handle by a university laboratory (in [13], Birth of an Electronic Computer, page 62.)
Vacuum tubes and transistors were expensive, and ferrite cores were dirt cheap is the perception of Goto and others.
Structure and Principle of Parametrons
Two donut-shaped ferrite cores were each wound with a wire of the same number of turns.
Two cores were required. The wire turns for two cores were configured
in opposite directions, i.e. opposite directions in two cores, for exciting current and oscillation current.
This is to isolate the primary and secondary parts: for example,
it cancels the DC bias that changes the magnetic property of core.
DC does not reach the oscillation part with the opposite twisting.
Note:This reversed turn is hard to recognize in figures at casual reading, but
all the figures quoted try to stress this using various visual cues,
but it is still subtle and easy to miss.
Figure 1 Parametron (Source: Information Processing Society of Japan)
Caption: Cores are visible on the vertical thick wires on the left
and right (external exciting current wire), and the thin wires are
signal wires.
When the resonant frequency of the circuit, composed of ferrite cores
and parallel-connected capacitors, is f/2, then we apply
external oscillation of frequency f to the ferrite core.
These ferrite cores were connected in series. A single capacitor was connected
to form a resonant circuit. An excitation wire passed through the
core's hole, and when alternating current (precisely, a combination of
direct current that changed the inductance of the core and alternating current) flowed through it, the
magnetism of the ferrite core caused the resonant circuit to oscillate
due to parametric excitation.
Separate wires may be used for direct current and alternating current
although most of the figures and diagrams in this application document
seem to use one wire.
Oscillation at half the frequency of the original external vibration
is amplified and observed.
In Goto's parametron, information can be stored by correlating the logic
states 0 and 1 with the difference of the phase of induced
oscillation.
There are two modes of oscillations that has the phase, exactly π apart.
(So parametron has the inherent capability to keep a bit of information.)
Parametric oscillation can be found in many systems. For example, to
move a swing in a park, we use parametric oscillation. The length of the rope
or chain used to suspend the seat of a swing, or more precisely the
distance between the center of the mass of the body plus the seat
material and the horizontal bar from which the swing is suspended is a parameter of the
oscillation. By moving our position up and down, we can change this
distance of the center of the mass, thus changing the parameter of oscillation.
By moving our body up and down once during a swing one way
(this means we move our bodies TWICE up and down during the full one swing back
and forth), we can excite the natural frequency of swing.
The frequency of excitation (our body movement) is twice the frequency
of swing.
[Remarks] A matchstick is included in the photo for size comparison.
Photo 3 Parametron in PC-1 Computer (Source: Information Processing Society of Japan)
As described, in addition to the property of stabilizing into two
clearly distinct states, the parametron exhibits an amplification
effect where differences in the initial state during excitation
determine the phase outcome. This is used for logical operation using
majority logic.
The outputs from parametrons to another parametron where a new oscillation was to be started can decide the 0/1 state (the phase difference in the final stable oscillation) by majority logic, i.e., analog addition of the signals. With parametron that uses the phase of the oscillation to distinguish 0/1 value as digital signal, logical operations called "majority rule logic" can be performed. Goto and others developed the full theory of this majority rule logic based on parametron to design logic circuits. (More about the theory of majority logic in the Obstacle to overcome section.)
While parametron was developed through theoretical and experimental efforts, it was very fortunate that the very first ferrite cores used in the first experiment for parametron was a copper-zinc type, developed by Yogoro Kato and Takeshi Takei, which later turned out to be verified to be best suited for parametrons among comparable ferrite cores available at the time. Although manganese-zinc and nickel-zinc ferrite have superior properties for other applications, copper-zinc ferrite proved best for parametrons [5] [Note 8].
To facilitate large-scale use in computing machines, smaller cores were preferable for reduced power consumption, leading Tokyo Denki Kagaku (today's TDK Corporation) to manufacture cores with a diameter of 4 mm. Later, the PC-2 used a "binocular type core" specifically designed for parametrons.
Diagram: Parametron called as shaped like binoculars marked with red rectangle.
4 φ core was the standard parametron initially.
2 φ core was used for memory.
Quoted from p.88 of Jiro Futami and Ryuji Shiozawa, "Parametron",
Hitachi Review, Feb 1960, vol 34. p.88, Figure 1 and Figure 2.
A circuit diagram that uses the binocular type core is described in
Appendix VI.
Parametron circuit
Parametrons are digital logic circuits that use nonlinear
property of circuit. They were applied to the realization of logic circuits.
They are characterized by high fault tolerance and low energy consumption in comparison to vacuum tubes.
Basic Structure and Operation of Parametrons
Parametrons are typically designed as two-terminal elements and are primarily constructed using capacitors and transformers. (There is also a resistor. See Appendix VI if you are curious electrical engineer.)
These elements transition between different "phase states," with the internal state changing in response to an external driving signal (input signal). Specifically, the amplitude and phase of the initial input signals, using by majority rule in the form of analog addition, determines the parametron to hold two stable states (0 and 1) distinguished by the phase difference of exactly Pi and can be used perform logical operations. (See the Goto's seminal paper [9] or slightly redacted explanation in Appendix V for more details. This application document summarizes the pros and cons of parametron in contrast to vacuum tubes, mechanical relays, and transistors for historical significance and the exact details of the operation is relegated to the original paper [9].)
The parametron circuit has the following characteristics:
(a) Nonlinear operation:
Parametrons respond non-linearly to external input signals. (The sum of the initial input signals determines the phase of the amplified induced parametric oscillation.) This property enables their use in majority logic circuits.
(b) Stability:
Parametrons are quite resistant to external noise and unstable signals, offering high fault tolerance.
(c) Low power consumption:
Parametrons consume very little power during operation, which made them particularly attractive in electronic circuits of the early days of computing. (This low power consumption is RELATIVE to the power-hungry vacuum tube of that day. It consumed much more power than today's power-efficient devices, especially when one tries to run parametron at high speed. More about this in the Obstacle to overcome section.)
Counting Circuit Using Parametrons
As a prototype of later computer, a simple counting circuit was developed using parametrons [3].
When parametrons were used to create the counting circuit, the design focused on the following key features:
(a) Counting circuits itself:
The parametron-based counting circuit combine multiple parametron elements to implement counting.
(b) Creation of Digital logic circuit:
Parametrons can serve as basic logic gates (e.g., AND, OR, NOT), which can be combined to create more complex computational circuits, such as adders and multipliers. But do note that parametron operates on the majority rule logic, not on the simple Boolean algebra. (Being unable to use Boolean algebra for parametron initially posed a slight difficulty in logic design. See Obstacle section for more detail.) Majority logic can mimic Boolean logic. So it can build any complex boolean logic circuit. What is more, sometimes, the majority logic simplifies the circuit very much: for example, the arithmetic unit that handles carry propagation quickly used the majority logic circuit to its advantage.
The success of the counting machine that used counting circuit based on parametron encouraged the people at Takahasi laboratory and thus the construction of fully programmable stored-program computer PC-1 began.
The Birth of Parametron Computer: PC-1
The PC-1 (Parametron Computer No.1) is a fully programmable stored-program computer for scientific computing, assembled in the Takahasi Laboratory, Faculty of Science, at the University of Tokyo. Using a total of 4,200 parametrons, production began in September 1957, and the first computation was performed on March 26, 1958.
Note: A stored program computer had been created at NTT's Musasino Laboratory using parametrons in March 1957. One year early than PC-1. It was called MUSASINO-1, and had only 32 words memory initially because the memory device for MUSASINO-1 came late. 32 words were not enough for meaningful software development. The memory was expanded to 256 words one year later about the time PC-1 began full operation. MUSASINO-1 was a handmade prototype and experienced many hardware failures. Thus, it caused significant maintenance burden. It was used only inside NTT's research. This usage pattern was unlike PC-1 which was used widely by researchers in and outside the University of Tokyo as explained later. In this manner, MUSASINO-1 could not have much impact outside NTT whereas PC-1 had a big impact on Japanese computer scene. (More information and links on this MUSASINO-1 and billing machine in Appendix I.)
Photo 4 PC-1 Parametron Computer (Source: Information Processing Society of Japan)
[Remarks] Dr. Eiichi Goto (Left) and Prof. Hidetosi Takahasi (right) in front of Parametron computer PC-1
Photo 5 Adder using parametron (Source: Information Processing Society
of Japan)
[Remarks] A vacuum tube is included in the photo for size comparison.
Subsequently, multiplication and division circuits were added, and
finer adjustments were made to each part. Materials such as
parametrons, exciters (the circuit that adds external oscillation to
parametrons), and input/output devices were borrowed from the
following companies.:
- Toyo Soda Industries,
- the International Telegraph and Telephone Company (today's KDDI),
- the Parametron Research Institute,
- Japan Electronics Instruments, etc.
Additionally, measuring instruments were borrowed from the Telecommunications Research Laboratory of the Nippon Telegraph and Telephone Public Corporation, and some were purchased with the Asahi Science Grant.
Such arrangement was necessary for the poor status of university laboratory back then.
It took a total of about 300 person-days to assemble the computer, with the parametron circuit wired by a worker experienced in wiring relays at Fuji Tsushinki (today's FUJITSU). The assembly and adjustment of the magnetic core memory device were performed by staff from Tokyo Denki Kagaku (Today's TDK Corporation).
The parametrons used in the PC-1 were of the early type, produced at different times and with irregular characteristics. Despite the challenge of assembling all circuits by hand in a university laboratory, there were no significant failures in 1958, except for issues with the exciter and the input/output device that used vacuum tubes.
Notably, the AC magnetic core memory device was surprisingly stable and reliable in comparison to similar devices. One reason for this stability was that, despite using 41 vacuum tubes in total, the main memory system adopted a circuit that applied error-correcting code to its operation, ensuring functionality even if one vacuum tube failed.
The performance figures of the PC-1 is detailed in the attached table. It used a binary representation internally. (Note that there were transistor-based computers that still used decimal digits representation internally at the time.) PC-1 was a fully programmable stored-program computer with fixed-point arithmetic only.
Wide Usage for Scientific Research and Training
In 1958, PC-1 was the fastest computer in Japan. *1 It was the only electronic computer available to researchers at the University of Tokyo. Thus it was used frequently by the researchers there and from other universities.
*1: The fastest Japanese computer in terms of the speed of addition and multiplication by March 1958 (when PC-1 became operational) was FUJIC developed at Fuji Photo Film in 1956 using vacuum tubes. It had faster addition and multiplication instructions than PC-1. However, FUJIC was short-lived. At the time Takahasi, Goto et al wrote "It (PC-1, proposer's comment) is the fastest computer in Japan at this time..." in [5] (September 1958), FUJIC had been mothballed to be donated to Waseda University in September of that year. So FUJIC was not operational. FUJIC had maintenance issue because 2-3 vacuum tubes needed to be replaced each day [15]. This was typical maintenance burden of vacuum tube computers of reasonable size in that era. (Replacement was not cheap. Vacuum tubes were expensive!) Fuji Photo Film had used FUJIC for optical lens design. Once Fuji Photo Film decided to stop creating lenses, FUJIC was moved to one of its subsidies and then donated to Waseda University eventually. So PC-1 *WAS* the fastest operational computer of the time in Japan. (See Appendix-V for more on the speed comparison and the history of FUJIC.)
In 1958, the PC-1 operated for 9-12 hours a day, with 5 hours dedicated to various numerical calculations for the Faculty of Science at the University of Tokyo. The remaining time was used for program research in the Takahasi Laboratory. From October 1958, about 10 hours per week were devoted to practical training for students.
Considering that PC-1 was maintained by graduate students at Takahasi
laboratory, being able to run for close to half a day each day was a
remarkable achievement showing the stability of parametron-based digital circuit.
In addition to using parametrons, the PC-1 had the following features [5]. (PC-2 built in 1960 is the successor of PC-1.)
Table 1 PC-1 and PC-2 Features
The characteristics of PC-2 (eventually built in 1960) in the table were according to the first design plan drafted in 1958. [5]
High-Speed Memory Device: AC magnetic core memory
For PC-1, an alternating current (AC) magnetic core memory device was used, a method invented in the Takahasi laboratory. Alternating current means the use of AC signal to read the data stored in core as the direction of magnetization. It used non-destructive read method as opposed to the direct-current destructive read method common to the core memory device used elsewhere in the world at the time.
The use of AC current for reading made the output quite friendly to the input to parametron device because the read signal from the core memory device was an oscillation wave, friendly to parametron, unlike the DC current reading method.
By finding excellent magnetic
material suitable for this method through the help of TDK and using a
address selection circuit that applied an
error-correcting code at runtime, the device achieved exceptional stability and
reliability. The storage device was relatively small and reported at
the symposium on electronic computer storage in the fall of 1957 and
the Annual Meeting of the Four Electrical Societies of Japan in May 1958.
Arithmetic Circuit
The arithmetic circuit employed a fast carry propagation circuit that handled carry separately. It was basically a carry lookahead circuit in today's parlance (maybe better called carry selection circuit that handles the carries in O(sqrt(N)) logic step or time where N is the bit length of the number, and theoretically optimal), but back then there were some variations, and serious research was going on. Goto et al came up with their own implementation. The carry handling logic (carry selector) of PC-1 took advantage of the majority rule logic of parametron to implement such circuit in a better manner than mere boolean logic circuit. Some parametron elements in it accepted five inputs. Careful selection of the ferrite cores of the parametrons that generated the output fed to the five input parametron was necessary to make sure the majority logic worked as expected by tuning the input signal levels. (Thanks to the comment from a reviewer.)
Additionally, a control system was adopted that simultaneously managed
ongoing calculations and those to be performed next. This was basically a pipelining at a very shallow depth of two. This improved
the calculation speed of the parametron computer by 2 to 3 times
compared to systems not using these methods, without significantly
increasing the number of parametrons required. In 1958, the pipelining
and other concurrent operation methods were being studied in many
countries and could be applied to electronic computers using any type
of logic element, not just parametrons.
Interrupt
Later in 1959, an interrupt circuit was installed in PC-1 so that the event from
the attached tape reader "interrupted" the on-going execution of an instruction.
When this happens, further interrupt was prohibited by a flip-flop,
and then the next instruction address is saved into address 510, and a pre-installed
interrupt handler whose address is in address 511 is invoked
(actually basically control jumps to the address stored in 511).
The flip-flop that inhibited the further interrupt was reset at the
exit of the interrupt handler.
In this manner, the pending input from tape reader could be handled,
e.g., put into a ring buffer that is accessed both by the main program
and the interrupt handler.
This is basically the operation called later multi-tasking.
PC-1 was one of the earliest computers to realize multi-tasking using interrupt.
Plan for PC-2
In 1958, Goto and others began planning to build a more powerful machine based on the experience and results of the PC-1. The machine was named PC-2. Although the computation and storage methods of the PC-2 were not substantially different from those of the PC-1, the plan was to increase memory capacity as shown in Table 1, provide a high-speed I/O device, and add a floating-point computing unit and an address translation mechanism.
This time, however, the construction was solely assigned to an external company.
PC-2 was completed in 1960: Funded by the Ministry of Education, it was jointly developed with Fujitsu. It was later commercialized as FACOM 202. (More information on this PC-2 and FACOM 202 in Appendix I.)
PC-2 was delivered to Takahasi laboratory. There, PC-2 continued the PC-1's tradition of wide usage by scientific researchers from all over Japan.
A commercial version of PC-2, FACOM 202, was delivered to the Institute of Solid State Physics of the University of Tokyo, and TOYOTA. At the Institute, it was used to calculate band energy of solid and contributed to the world class research. (There is a short description of how FACOM 202 was used at the Institute of Solid State Physics in the newsletter of the Institute, p. 17, No. 5, Vol 4, December 1964, in Japanese, available online at https://www.issp.u-tokyo.ac.jp/maincontents/docs/tayori/tayori04-5.pdf)
PC-1 as a research vehicle of parametron performed very well. It was
finally disassembled when its power unit was rented to an exhibit done
by physic students during an annual open house event at the University
of Tokyo in May 1964.
Historical Impact of Parametron on Computers and others
The ability to build computers with a greatly reduced number of vacuum tubes and transistors led to the creation of many parametron-type computers in Japan at that time. Compared to relay-based systems, parametrons were faster and had no mechanical contacts that were the source of many failures, offering great advantages.
Despite the relatively short period of time while parametron enjoyed success, these computers were sold in large numbers the market. A detailed list of these commercial computers is in Appendix I. There were many. Certain model was a best seller of the time, selling more than 800 units in total.
Back in 1959 when the seminal paper of Goto [9] was published ("The Parametron, a Digital Computing Element Which Utilizes Parametric Oscillation", Proceedings of the IRE, Volume: 47, Issue: 8, pp. 1304 - 1316, August 1959), Goto mentioned the following. (Quote from the paper).
In 1954 the author discovered that a phenomenon called parametric oscillation, which had been known for many years, can be utilized to perform logical operations and memory functions, and gave the name "Parametron" to the new digital component made on this principle [1], [21]–[23].(Proposer's note: the reference numbers are original paper's.). A digital computing circuit made of parametrons may consist only of capacitors, ferrite-core coils and resistors, while diodes and rectifiers may be dispensed with. The parametron, therefore, is considered to be extremely sturdy, stable, durable, and inexpensive. Owing to these advantages, intensive studies have started in several laboratories in Japan to apply parametrons to various digital systems. AT PRESENT, NEARLY HALF OF THE JAPANESE ELECTRONIC COMPUTERS IN OPERATION USE PARAMETRONS FOR LOGICAL ELEMENTS. (The EMPHASIS is by the proposer.) Further applications have been made to such devices as telegraphic equipment, telephone switching systems and numerical control of machine tools.
However, the rapid performance improvement of junction transistors, which became mainstream shortly afterward, outpaced parametrons in operating frequency. Transistors had broader applications, such as in radios and analog circuits also, while parametrons were dedicated to the use as logic elements. Thus the investment on R&D of transistors far outweighed that of parametron. By the mid-1960s, parametrons were almost entirely replaced by transistors and fell into disuse.
Non-computer applications of parametron
This application document so far focuses on digital computer applications of parametron. But there were digital circuits built using parametrons [13]. Here is a brief summary. In a sense, these are the devices where parametron logic element with ease of use and design, and durability shined.
- KDDI (its predecessor, KDD) built a machine to convert Morse code to Teletype character code using parametron. This converter was used for transmitting news messages from the Melbourne Olympic Games in 1956. It performed flawlessly without breakdown. Note that it was built in less than two years after parametron was invented.
The success of the converter encouraged KDDI to create further devices such as re-generative repeater for long distance transmission (to re-shape the received signal for relaying purposes), and automatic repeat-request machine (ARQ machine) for error correction/recovery of TELEX communication. The ARQ thus developed using parametron was the world's first electronic ARQ device and many ARQ units were exported. So parametrons WERE USED OUTSIDE JAPAN although the users were unaware of parametron usage. The first ARQ using parametron was created in 1956, only two years after parametron was invented. A written record exists to show that it was still in use by KDDI in 1977 at the latest, more than twenty years after its birth. It shows the robustness of parametron-based device. (More about these devices developed by KDDI in the entry of KDDI in Appendix I.)
- Numerical control of machining devices were also built using parametrons. Numerical control was used to position the machining tools precisely to create desired shapes. Before that, the machining devices mimicked the desired movement using curved template. Creating the curved template was a very time consuming task and the resulting accuracy was not quite good for modern aerospace industry, for example. Thus numerical control of machining device, initially developed and popularized by MIT researchers, was a hot topic circa 1950s and early 1960s. Numerical control device was researched and built in Japan. A device is mentioned in Hitachi's entry in Appendix I since Hitachi collaborated with the researchers at the Japanese government institute called Agency of Industrial Science and Technology to create the device in 1956. However, later devices deployed in the field used transistor and germanium and parametrons were not used.
Fujitsu also developed a numerical control unit that used parametrons in 1956 as a prototype for later deployment. But again, later models deployed and commercialized did not use parametrons. This device is discussed in Fujitsu's entry in Appendix I.
Technology landscape was changing rapidly around the year 1960.
- A very unique use was to detect the inter automobile distance to prevent car collisions. Wire loops as part of parametric oscillation were buried under the road. When a car approaches the loop, its loss of high frequency response and this was detected by the change in parametric oscillation strength and thus the distances between cars can be measured. (This is basically what is done today for car position detection although no one thinks of saying that we use parametron although Electromagnetic principle remains the same.) Unfortunately, not much information on this usage back in 1950s and 1960s is available online.
Some of the developed devices were used late into middle of the 1970s (according to a reviewer), and it is true that parametron-based ARQ by KDDI was still in use in 1977 according to the record kept at a government website. (See KDDI entry in Appendix I.)
Nurturing the new generation of computer engineers and users
Parametron's impact was also human resources and science computing via the parametron computer, PC-1. Many future computer scientists/engineers were born among the early users of PC-1, the stored-program parametron computer at the University of Tokyo. Many of them were graduate students there and at other universities. Also, many researchers created scientific computation library routines.
Only very sketchy information in the early days of PC-1 computing remains today. But there is a record of a seminar on programming organized by the Japanese Society of Physics in 1959. This is possibly one of the first such computer training seminar in Japan according to Takahasi [13].
That used PC-1 as the target computer to write programs. Below the agenda of the seminar is quoted.
The timetable of the seminar in 1959. Quoted from "TOYOTA and parametron electronic computer FACOM 202", Yoshihiro Ishibashi, available online at: https://www.toyotariken.jp/_/media/page/about/research-report/pdf/Toyota-Report_No.77_62.pdf
The seminar was held from Aug 31 to Sept 7 in 1959. Below the titles (translated in English) and speakers of the lectures are quoted. (The original printed agenda is scanned and OCR version is available in [14] IIJLAB 2008, パラメトロン計算機 PC-1 1958-2008 パラメトロン計算機記念会, "Parametron computer PC-1 1958-2008" by Parametron computer anniversary committee in Japanese, https://www.iijlab.net/~ew/pc1/pc150th.pdf)
This shows the impact PC-1 had on Japanese academia, in this case to physical sciences.
The title of each lecture and the name of the lecturer is given below.
Introduction to Electronic Computing by Hidetosi Takahasi
Arithmetic Instructions for the Parametric Computer PC-1 by Eiichi Goto
Experience in using electronic calculators by Takehiko Shimanouchi
Applications to Geometric Optics, etc. by Bunji Okazaki
Application to Crystal Analysis by Yoshio Takeuchi
Numerical Analysis for Computers I (Linear Computation) by Shigeichi Moriguchi
Numerical Analysis for Computers II (Numerical Integration and Differential Equations) by Ayao Amemiya and Masataka Ariyama
How to make programs (flow charts and their examples) Yoshihiro Ishibashi
Application to Meteorology by Kikuro Miyakoda
Application to Fluid Mechanics by Isao Imai
How to program (how to use subroutines) by Takashi Soma
How to make programs (how to make tapes using R0 R1) by Keisuke Nakagawa
Electronic Computing in Universities I by Yonezo Morino
Electronic Computing in Universities II by Mitsuro Omori and Shigetoshi Katsura
Monte Carlo Method by Yoichi Fujimoto and Eiichi Goto
Applications to Quantum Mechanics by Masao Kotani
Applications in OR and Control Engineering by Koh Hosaka
How to find errors in programs by Eiiti Wada
Future computers and programming by Hidetosi Takahasi
Panel Discussion "Current Status and Future of Computers"
Chair: Takahiko Yamamouchi
Panelists Takashi Isobe (The Univ. of Tokyo), Koh Hosaka
(Technical Research Institute of the Japanese National Railways),
Hidetosi Takahasi (The Univ. of Tokyo), Shigeichi Moriguchi (The
Univ. of Tokyo), Masao Kotani (The Univ. of Tokyo), Hiroshi Wada
(Electro-technical Laboratory, ETL), Zen'ichi Kiyasu (NTT),
Takeshi Kayano (NTT)
Many names in the list of lectures would hold important positions
later in the computer industry in Japan or academia, both in software
and hardware area, and physical sciences and engineering. PC-1's legacy lived on after parametron fell into
disuse.
The list of available routines and the type of calculations
performed on PC-1 is in Appendix II.
Popular Culture
When many people in the first generation of computer industry and in academia in Japan got their first taste in programming on parametron-based computers, it had left its imprint on education as you can see from the number of parametron computers still displayed at many education institutes and research facilities today in Japan. (See Appendix I for such examples.).
Parametron also left a legacy or urban-legend memory on the mind of early electronics engineers.
Thus, it has even popped up in very popular comic lately.
"Dr. Stone" is a very popular Japanese comic (and turned into animation) with total circulation more than 18 million copies. (See Wikipedia: https://en.wikipedia.org/wiki/Dr._Stone )
The story is an SF-like one.: A mysterious catastrophe petrified all human beings and after a few thousand years, one of them somehow was revived to life and investigated the situation, and tried to re-create the human civilization with some partners he found.
Along the way, they had to create many instruments that one took for granted in everyday life. During the course of urban planning, the characters wished for a calculator and computer. Vacuum tubes are way off their manufacturing skill to say nothing of the semiconductors that required purified semiconductor crystals (in the SF story timeline, semiconductor would be available in 10 years' time, but the calculators are wanted NOW).
Then, parametron to the rescue (!). Ferrite cores could be manufactured by heating and pressing metal powder. Thus in the comic, 0.2 million parametron cores were going to be created using such crude method so that they could be used to create computers and calculators.
The above drawing is from the 206th installment of Dr. Stone (August 2021).
There is an enlarged image of a ferrite core in the upper left.
Ferrite core is created by burning the powder of metal material in
the lower left.
You can read the original pages from officially sanctioned sources such
as the following (in Japanese). https://www.mangajikan.com/chapter-104565.html
Note: The above drawing is quoted as fair use for academic purposes of the copyright material.
(See Shueisha publishing's copyright stance (in Japanese).
https://faq.shueisha.co.jp/faq/show/32?category_id=8&site_domain=default )
The proposer thinks there is a slight abnormality in the
enlarged core shape in the drawing (upper-left).
(See the enlarged photo of a rare remaining PC-1 board in Appendix VI.)
Still having this comic is a great way to teach
parametron existed to the youth today, and tell the general audience
that the principle behind parametron is still viable as Adiabatic
Quantum-Flux-Parametron (AQFP) (see "Quantum flux parametron" in Appendix III, "Parametrons in
Disguise") in the
21st century.
What obstacles (technical, political, geographic) needed to be overcome?
Obstacles (Technical, Political, Geographic) Needed to Be Overcome
The parametron, invented by Dr. Eiichi Goto in 1954, is recognized as
a significant milestone in the history of logical device for electronic computing. This pioneering technology faced numerous obstacles on its path to success, spanning technical, political, and geographic challenges.
Technical Obstacles
Material Limitations
During the early 1950s, the materials available for electronic components were limited. The parametron, which relied on the parametric excitation of non-linear inductance to achieve switching, required high-quality inductors and capacitors. The scarcity of high-quality materials and the limitations in manufacturing technology posed significant challenges. Achieving the necessary precision and reliability in component fabrication was a major hurdle.
As Goto mentioned later, he was lucky to use the right ferrous material as far as the ferrite core, the essential part of his parametron device, was concerned. He picked up a ferrite core very suited to parametron in his first experiment. Had he not used the particular material, the search for the right material alone may take a couple of years, thus missing the opportunity to fill in the gap between vacuum tube and transistor with very reliable parametron device.
Design Complexity
The initial physical design of parametron was intricate and required a deep understanding of non-linear dynamics and resonance phenomena. Goto analyzed the parametric oscillation by modeling the excitation using so called Mathieu's differential equation [9]. This complexity made the design and construction of practical parametron circuits challenging. Researchers had to overcome the difficulties of designing circuits that could maintain stability and reliability under varying operational conditions.
But, once the physical circuit was designed properly with well selected proper material, parametrons performed very reliably.
This was the key reason the stored-program computer built at a university laboratory, PC-1, operated so successfully without any dedicated operator. Serious breakdown of the PC-1 computer did not occur often, The graduate students and the researchers there helped the external science researchers who ran long-running computer programs. Programs could run for a few hours unattended on PC-1, which was a feat in itself in its time.
Another key challenge in parametron design was its reliance on
majority logic, which differed significantly from the binary Boolean
logic prevalent in electric/electronic circuit design at the time and
today. Boolean algebra, designed for binary AND, OR, and NOT
operations, was inadequate for majority logic circuits.
To address this, Goto and Takahasi developed a novel approach to majority logic for Parametron-based designs.
First, they introduced a graphical representation to visualize the majority logic employed in Parametron circuits, making the interaction of majority logic more visible. The interested readers are invited to read the Goto's seminal paper where the notation is explained.[9] It is a very interesting exercise to implement various logical operations using majority rule logic of parametron. A few examples are given (taken from [14]).
Figure - Majority rule circuit representation of Parametron
A circle stands for one parametron element. It can accept multiple input signals. The +/- in the circle indicates an input signal that is constantly set either to 1 (+), or 0 (-). The bar in the connection line stands for negation. In parametron, negation does not need a logic element. Reversing the direction of winding through the ferrite core achieves that.
The upper left is a Boolean AND operation. Since there is a constant 0 input (note the "-" in the circle), only when two other inputs are both 1s, the majority rule produces output 1, thus implementing Boolean AND operation.
The example at the middle in the left is boolean OR operation. Since there is a constant 1 input (note the "+" in the circle), if at least one of x or y is 1, the majority rule produces one as output, thus implementing Boolean OR operation.
The lower right is a full adder. x is a carry from
the addition of lower binary digits. The circuit produces the sum, 's', of x, y, and z,
and produces 'c', carry, that is needed to pass to the next higher bit
calculation.
Building on this, they systematically analyzed all possible four-input majority logic configurations [13], optimizing circuits to achieve desired outputs.
Their breakthrough was the classification of majority logic circuits into self-dual equivalence classes, which groups equivalent functions to reduce design complexity ([6], “Some Theorems Useful in Threshold Logic for Enumerating Boolean Functions,” E. Goto, H. Takahasi, IFIP Congress, 1962; see also “Classification of Ternary Logic Functions by Self-Dual Equivalence Classes,” T. Soma, T. Soma, 41st IEEE International Symposium on Multiple-Valued Logic, 2011. The latter is an extension of parametron to use three valued logic, using three different phases instead of two).
This classification provided critical insights for designing logic
circuits using Parametrons and other threshold logic devices. It
remains relevant today, enabling efficient designs for quantum flux
parametron (QFP) and its adiabatic version, AQFP, which have gained
significant attention, as discussed in the “Features” section.
Thermal Management
The operation of the parametron at fast speed involved significant energy dissipation, leading to heating issues. (See Appendix VI for detailed discussion of heat dissipation, etc. for electronically inclined.) Effective thermal management was crucial to ensure the longevity and reliability of the Parametron circuits. Designing cooling mechanisms and optimizing the thermal performance of components were critical technical challenges.
In the case of the PC-1 computer mentioned in this submission, it did not push the speed limit much. Thus, this computer at the University of Tokyo, employing approximately 4200 parametrons did not suffer from catastrophic heat failure although it did not have an active cooling mechanism (except for fans).
According to a memoir by Keisuke Nakagawa (in "パラメトロン計算機 PC-1 1958-2008 パラメトロン計算機記念会,", Parametron computer PC-1 1958-2008" by Parametron computer anniversary committee in Japanese, https://www.iijlab.net/~ew/pc1/pc150th.pdf), who was a graduate student at Takahasi laboratory where PC-1 was placed, PC-1 was used as follows.
"PC-1 came to be used by researchers of the Faculty of Science for their research. The computer time was made available even during night hours. PC-1, despite the meager 512 words storage (one word was 18 bits), allowed scientists to do variety of computations in many fields. PC-1 operated in a room without a special air conditioner for it. People opened windows during summer, but closed the windows with steam central heating running during winter. But PC-1 kept running demonstrating parametron's stability to the world. However, if PC-1 showed flaky behavior, available graduate students followed the predefined recover steps so that PC-1 ran again." (in submitter's translated summary from the original Japanese).
Mr. Nakagawa was a graduate student back then.
Integration with Existing Technology
The Parametron was a novel technology that needed to be integrated with existing computing systems and peripherals. Ensuring compatibility with the infrastructure of the time, including input/output devices and memory systems, required innovative solutions. Researchers had to bridge the gap between the new Parametron technology and the established electronic computing landscape.
These are the issues any new logic device technology faces, so not a particular obstacle specific to parametrons.
PC-1 computer that was built with parametron was connected to a paper tape reader, and teletype writer device and so the interfaces with simple I/O devices were available.
Main memory was developed specifically for PC-1, namely the AC-drive
core memory system that produced its output (1/0) in terms of
oscillating signal which was very friendly to parametron that could
accept such input directly.
Miniaturization
Early electronic components were bulky, and miniaturizing the parametron circuits was a daunting task. Reducing the size of the parametron while maintaining its functionality and performance required advancements in component design and fabrication techniques. This miniaturization was essential for making the parametron commercially viable and practical for real-world applications.
Binocular type parametron core mentioned in the Historical impact section was an effort for the initial miniaturization path.
Unfortunately, obviously, in the long run, there was no chance for ferrite core parametron to compete with the transistor in the miniaturization race.
However, later in the 1990s, miniaturized superconducting device was used
to create a parametron-like
behavior, and is a very hot topic now that the low-power consumption
of AQFP attracts attention in SDGs age. (https://en.wikipedia.org/wiki/Quantum_flux_parametron,
Adiabatic Quantum-Flux-Parametron: A Tutorial Review
https://www.jstage.jst.go.jp/article/transele/E105.C/6/E105.C_2021SEP0003/_pdf/-char/en
Political Obstacles
Cold War Era: The development of the Parametron occurred during the
Cold War, a period marked by intense geopolitical tensions between the
Eastern and Western blocs. This political climate influenced research
priorities and funding allocation. Gaining support and resources for
parametron research in Japan, which was rebuilding its economy and
technology sector after World War II, was a significant challenge.
Funding and Resources
Securing funding for innovative research was a constant struggle in post-war Japan. Government and institutional support were limited, and researchers often had to rely on private industry partnerships and international collaborations. Convincing stakeholders of the potential benefits and applications of the Parametron required substantial effort and persuasion.
In [7] ("Some Important Computers of Japanese Design", IEEE Annals of the History of Computing, Vol.2, No.4, pp. 330-337. Oct.-Dec. 1980), Professor Hidetosi Takahasi remarked in its abstract to describe the era thusly.:
"we were on the verge of starvation in the ruin of our defeated country", ... "We were starved for knowledge as well as for food".
Such was the atmosphere of post-war Japan when parametron was invented.
Goto himself and the laboratory of Professor Takahasi back then was lucky to obtain a few big corporate backing after initial report of Goto on the principle and experimental result of parametron device caught the eyes and ears of the people of a big research laboratory, Musasino Laboratory of NTT, a large telephone operator Kokusai Denden (KDD, precursor of today's KDDI), and others. This backing helped parametron take off.
Still the budget was small in comparison, say to MIT's: Goto mentioned the following in a magazine interview (in Japanese). The following English translation by the submitter is an excerpt from the interview in Japanese.: https://ascii.jp/elem/000/001/221/1221954/3/
——How was the reaction to the PC-1 parametron computer?
“It wasn't that big a deal. *1 Overseas, there were already things like ENIAC and EDSAC ten years earlier. Also, the processing speed of the parametron was slower than that of transistor computers. The clock speed of transistor computers was 1 megahertz, whereas the parametron computer was at most 10 to 30 kilohertz.
Much later, I became friends with McCarthy from MIT, and he told me, “The parametron is an interesting idea, but why did you make such a slow element?” He could say that out of ignorance, but our budget was only about one thousandth of MIT's *2. But I did feel a sense of achievement in having made computers available at the university at that time. Even though it only had 256 words of memory, lots of people came to use it. There weren't any other computers around at the time.
*1: The submitter thinks Goto underrated PC-1 much as its creator because of his modesty and because he was too aware of the speed issue. As he mentioned, PC-1 became immensely popular among the academic researchers, and many commercial entities adopted parametron to create very successful commercial computers. In Goto's mind as its leader to build PC-1, the computer may not have had that much impact, but others and commercial enterprises in Japan of that era had a very different opinion.
*2: Obviously, Goto referred to the budget of similar computer project, NOT the entire MIT budget.
Many private companies adopted parametron to build computers and calculators soon. Some models sold very well in the era. (See the list in Appendix I).
Also, after parametron became famous, a non-profit organization OUTSIDE
the University of Tokyo was established to handle the funding and
intellectual property issues, which lead to the next item.
Intellectual Property and Collaboration
Navigating intellectual property rights and fostering collaboration with international researchers were political challenges. The exchange of knowledge and technology between countries was often hindered by political considerations and restrictions. Establishing frameworks for collaboration and ensuring the protection of intellectual property were critical for advancing Parametron research.
Because the significance of parametron was so clear to the early adopters, they began helping Goto patent the inventions. To proceed with international patenting, an outside NPO called Parametron Research Laboratory (tentative English translation) was formed in Mar 8, 1957 (The birth of a computer, 1971 [13]) and the intellectual property matters were handled by this entity after that, freeing Goto, Takahasi and others at the University so that they could focus on technical inventions at hand.
Two US patents for Goto's parametron were granted.: US Patent 2,948,818 [P1] and [P2] US Patent 2.948,819. They were initially turned down with a comment, "it does not operate". ([13] p.82]). A US company, NCR licensed the patents eventually but did not produce anything after about a year and its interest seemed to have disappeared. It was too late, in a sense, since the transistors became more robust, and parametron's advantage was disappearing very fast.
Goto's mentor, Professor Takahasi wrote the following in 1971 [13] (in submitter's English summary). This is a food for thought for today's inventors/researchers.
"Fearing the publication of the technology might invalidate the patent applications, we took the hush hush approach, not publishing the new technical results any longer. But with hindsight, it may have been better to adopt a more open approach to share technology even its current problems, with other parties early. Then the technology might have been used wider (Proposer's comment: outside Japan, too, implicitly) and problems may have gotten solved with more inputs from many parties. I think this approach might have worked better for parametron."
Geographic Obstacles
Research Infrastructure
Japan's research infrastructure was still recovering from the devastation of World War II. Establishing well-equipped laboratories and securing access to advanced research facilities were significant geographic challenges. Researchers had to overcome the limitations of the existing infrastructure and build new capabilities from the ground up.
Although Takahasi lab was hardly a rich laboratory back when parametron was invented, it enjoyed a better than average status because the University of Tokyo was the largest nation-run university of that time.
Access to Global Knowledge
Geographic isolation posed challenges in accessing the latest research and technological advancements from other parts of the world. Japanese researchers had to find ways to stay informed about global developments in electronic computing and incorporate this knowledge into their work. Building networks and establishing communication channels with international researchers were essential for overcoming this obstacle.
The following anecdote may seem outlandish to readers in the 21st century. However, Professor Takahasi (and people in other field such as physics, etc. of that time for that matter) mentioned that he learned of transistor and other technical discoveries in the world, via magazines made available at a library established by an occupying forces stationed in post-war Japan. University libraries were not functional at all back then.
There was a library in Hibiya, Tokyo, established by the general headquarters of the occupying force in post-war Japan and that was the place to go to read the latest American magazines including technology/science ones. CIE Information Center Library was it. So being in Tokyo, the capital of Tokyo was important. (There were similar CIE libraries in other parts of Japan, but Tokyo's Hibiya one seemed to have been largest. https://ja.wikipedia.org/wiki/CIE%E5%9B%B3%E6%9B%B8%E9%A4%A8) Takahasi, Goto and others seemed to have learned of the EDSAC computer which they seemed to have studied extensively before the construction of PC-1 from learning about it through reading at this library in Hibiya. The library no longer exists.
Goto himself did not leave much about his study style in writing and the submitter could not learn much about WHERE he obtained knowledge. His research style, though, was to think hard about a topic himself very much and arrive at a solution or two, or even more before seeing other people's previous work.
Market Acceptance
Introducing a novel technology like the parametron to the global market requires overcoming geographic barriers. Domestic market in Japan accepted parametron very quickly and produced computers based on it. Some of them (NEAC-1201 and NEAC-1210) sold more than 700-800 units, which was a big number for computer sales at the time. (See Appendix I - "Detailed list of commercial Parametron computers")
Convincing international markets of the parametron's advantages and securing adoption outside Japan involved addressing cultural and logistical challenges. Establishing distribution channels and support networks in different regions were crucial for the parametron's success.
That NCR in U.S.A. licensed Goto's parametron patents was a testament to the advantage of the parametron device at the time. Whether there was a strong support network to help NCR proceed is now a question of historical interest. There was no internet, no e-mail, no international fax.
There is anecdotal evidence that European companies also monitored parametron development. But it is only in one person's memoir of parametron. ("TOYOTA and parametron electronic computer FACOM 202", Yoshihiro Ishibashi, available online at: https://www.toyotariken.jp/_/media/page/about/research-report/pdf/Toyota-Report_No.77_62.pdf) Submitter's translation of the relevant paragraph follows.:
--- Yoshihiro Ishibashi's recollection
The author's specialty is “ferroelectric properties”, and when talking with foreign researchers who have the same specialty, the topic sometimes turns to graduate school days. When the parametron is mentioned in such a situation, of course it is passed over with “what's that?”, but there was one occasion when I got a reasonable response. It was from a British researcher of the same age, and therefore someone who knew about the situation with computers in the 1960s, who said, “I know about the parametron. Thomson (France) studied it, but it didn't work, did it? Did you really make a computer using parametron in Japan?” It seems that parametrons were not completely ignored in Europe. It didn't mean much, but it made me feel happy somehow.
Note: Yoshihiro Ishibashi was a graduate student in 1958 and studied under the supervision of Eiichi Goto. He was still a graduate student when he gave a talk at the programming seminar hosted by Japan Physics Society using PC-1 in 1959.
These technical, political, and geographic obstacles were significant, but the dedication and ingenuity of Dr. Eiichi Goto and his team led to the successful development and implementation of the parametron. Their achievements laid the groundwork for future advancements in electronic computing in Japan and demonstrated the resilience and innovation of the scientific community in overcoming complex challenges.
What features set this work apart from similar achievements?
Features that set this work apart from similar achievements
Comparison with Other Methods
In the following, we compare the parametron with vacuum tube, electromechanical relay, and transistor types of the 1950s:
Please note that advantage and disadvantage of parametron such as speed,
power consumption, etc. are discussed in the relationship with other
devices in the following.
For example, the power consumption of PC-1 *WAS* huge in
comparison to today's power efficient computers, but was definitely
smaller RELATIVE to the vacuum tube computers.
PC-1 computer's power consumption was 3KVA as shown in table-1.
Pros
(a) The price is significantly lower compared to vacuum tubes (and more so than transistors).
(b) Compared to relays, it can operate at higher speeds.
(c) It is more stable than vacuum tubes with relatively short life cycle and early transistors.
(d) Ferrite cores possess physical strength due to being made of ceramics.
(e) Errors due to radiation are less likely to occur.
Cons
(a) It consumes more power than transistors (less power than vacuum tubes).
(b) While transistor calculators of the same era achieved an operating frequency of 1 megahertz, the parametron operated at about 10-30 [kHz], making it slow.
The power consumption (a) and slow speed (b) were due to the fact that the parametron's internal frequency is half the excitation frequency, and that an observation of at least several full stable cycles of parametron output is necessary to ascertain the phase shift to determine the logical output, i.e., 0/1. Thus the frequency of parametron's logic operation at least 16-20 times as slow as the parametric oscillation frequency, which is half the excitation oscillation frequency.
To try to achieve higher speed operation of parametron, it was necessary to feed much very fast external excitation frequency. So this was the inherent limit of parametron speed. And it also compound the power consumption issue, (a) above, since the excitation frequency circuit is basically the power source, and it had to operate at a very high frequency and it had to offer three output with different phase at that (Detailed discussion of the exciting frequency and the operation frequency is given in in Appendix VI to avoid cluttering the main text.)
When Goto visited MIT later, and became friends with John McCarthy of Lisp fame there, he was asked, "Parametron is an interesting idea, but why did you make such a slow device?". (The true answer to this question was because the research budget of Goto's at the University of Tokyo was 1/1000 of similar MIT project according to Goto, but he did not say this to McCarthy then because of his national pride.) The speed difference between parametrons and junction transistors is notable; parametrons operated at a much lower frequency, leading to slower performance.
A detailed comparison of speed of various parametron computers and others in the same era is given in Appendix V.
Because ferrite core has hysteresis, when excited at high frequency, it loses the energy as dissipation heat, which leads to the next disadvantage.
(c) Due to the heat generation (caused by loss due to the hysteretic characteristics of the magnetic material), increasing the operating frequency causes the ferrite core to overheat, altering its magnetic characteristics (burning) and hindering operation.
(d) It does not function properly when miniaturized, making it challenging to integrate using micro-fabrication technology.
(e) Boolean logic used for other devices was not useful. Parametron used majority logic instead of simple Boolean logic of AND/OR/NOT. So the logic design was initially difficult until Goto's team that created PC-1 came up with the better classification of multi-input functions to give a bird view of the logic design that uses majority logic, so to speak. (See the full discussion in "Design Complexity", in the Obstacle section.)
(f) Fan out is not large. (This disadvantage was thanks to the comment from one of the reviewers).
You cannot drive many parametron devices from the output of a single parametron. It was observed that the parametron's oscillation is adversely affected if too many fanout signals are extracted.
So, for the design of PC-1, the fanout was limited to 12. We can always duplicate a logic stage to duplicate the desired output so that this fanout limit is not overcome and the speed is not lost due to the additional logic stage. Yet, this step may introduce an overflow of the fanout limit in the previous logic circuit stage. We have to repeat the process until the fanout condition is met at all the logical stages. Will this step eventually stop or will the fanout limit make the design using parametron infeasible?
Luckily, Goto proved that these steps to eliminate fanout overflow will always stop as long as fanout F is larger than fanin (I = 5 in the case of parametron used for PC-1). (E. Goto: "A Note on Logical Gain", IEEE Transaction on Electronic Computers" EC-13, October 1964) This fanout and fanin relation holds for any type of logic circuit.
So, although this fanout limit of parametron may have expanded the
area of circuit boards to accommodate more parametrons, this was NOT
an inherent design limitation.
Logic Elements Similar to Parametrons
Here we look at some devices similar to parametrons.
Capacitance Variable Type Parametron
Goto's parametron uses variable inductance (L). (Do note that his seminal paper [9] mentioned the use of variable capacitance (C) as well. He knew that he could create a parametron-like device using variable C as early as 1955 [16] or earlier obviously.) Around the same time as Goto's patent application (filed in April 1954), von Neumann proposed the idea of using parametric oscillations that change capacitance (C) or inductance. But his patent described the very meta-level of the IDEA of a parametron using microwave as the oscillation source and whether he tried to implement it and would have succeeded remains very mute. But Von Neumann did notice both the amplifying feature of input signal and memory feature of the parametron-like device as Goto did. (Von Neumann patent: U.S. Patent 2,815,488 "Non-linear capacitance or inductance switching, amplifying, and memory organs " applied in 1954 and published in 1957. Accessible online at the following URL: https://patents.google.com/patent/US2815488A/en)
There was a demonstration of C-type parametron by Dr. Hiromichi Hashizume et al in 2018. (https://siogadget.wordpress.com/2008/03/27/%E3%83%91%E3%83%A9%E3%83%A1%E3%83%88%E3%83%AD%E3%83%B3%E8%A8%88%E7%AE%97%E6%A9%9Fpc-1%E8%AA%95%E7%94%9F50%E5%91%A8%E5%B9%B4%E8%A8%98%E5%BF%B5%E3%81%AE%E4%BC%9A%E5%90%88%E3%81%A7%E3%81%AE%E5%B1%95/ in Japanese ). According to Hashizume (oral communication), he had tried to recreate Goto's parametron, but he could not obtain a ferrite core that showed favorable behavior, and thus he switched to variable capacitance variety instead.
Thin-Film Magnetic Material Parametron
Following the same principle, parametrons using thin-film magnetic
materials were also studied, but they were not commercialized on a
large scale as logic device.
However, it was developed into a wire memory based on the wire
parametron in 1963 by KDDI. It is described in KDDI's entry in
Appendix I.
Magnetic Flux Quantum Parametron
Magnetic flux quantum parametrons have also been studied. Proposed in 1984 by Goto (E. Goto, "Josephson pair elements", Proc. 1st RIKEN Symp. Josephson Electronics, pp. 48-51, 1984., E. Goto and K. F. Loe, DC Flux Parametron, Singapore: World Scientific, 1986.). as a switching element capable of high-speed operation up to 16 GHz using the Josephson effect, these elements have principles similar to those of parametrons. Goto mentioned in an interview, "The fact that the principle is similar to parametron means that the same person can think of it." In addition to high-speed performance, these elements are characterized by power saving compared to other superconducting devices (such as Josephson elements), but large-scale integration has not been achieved yet. While they use quantum mechanics, they are not considered quantum computation in the conventional sense. [ref: https://en.wikipedia.org/wiki/Quantum_flux_parametron]
Goto invented the QFP in 1984 as a low-power high-speed switch element. but the industry's mainstream investigated different superconducting logic circuit using different mechanisms since then.
Additionally, a reversible computational element using adiabatic quantum parametron (AQFP) approaching the limit based on Landauer's principle has been proposed.
Lately in the age of SDGs, the energy-saving feature of QFP has revived its popularity, and AQFP is now a hot topic. The history of QFP and recent trends are explained in detail in the following reference succinctly: Adiabatic Quantum-Flux-Parametron: A Tutorial Review, https://www.jstage.jst.go.jp/article/transele/E105.C/6/E105.C_2021SEP0003/_article
Appendix III discusses briefly the parametron-like devices including QFP that were born after the original parametron was invented by Goto.
Why was the achievement successful and impactful?
The invention of the parametron in 1954 was successful and impactful
due to several key factors.
Firstly, the parametron, developed by Goto, was a revolutionary advancement in electronic computing technology. Unlike traditional vacuum tube-based computers, the parametron utilized parametric oscillation, which significantly reduced power consumption and reliability was improved by leaps and bound. This technological advantage addressed the major limitations of early computers, making them more practical for widespread use.
Additionally, the parametron's design was versatile and adaptable, allowing for various applications in different fields. It found use in scientific research, industrial automation, and other field. The flexibility and efficiency of the parametron paved the way for further advancements in computing technology and set the stage for future innovations.
Lastly, the parametron's success was not just limited to its technical merits. It also had a lasting impact on the development of computer science education in Japan. Many universities and research institutions adopted computers that used parametron for teaching and research purposes, fostering a new generation of computer scientists and engineers. (Some universities display the original parametron computers thus used as memento of the birth of computer education at their institution. See Appendix I.) This educational impact contributed to the rapid growth of Japan's technological workforce and the country's continued leadership in the field.
The parametron's impact extended beyond just technological innovation. It played a pivotal role in establishing Japan as a player in the field of computing during the post-war period. At a time when Japan was rebuilding its industrial and technological capabilities, the success of the parametron demonstrated the country's ability to innovate and compete on the global stage. This accomplishment instilled a sense of national pride and confidence in Japanese scientists and engineers. This left an impact on the future computer industry.
In summary, the success and impact of the parametron invented in 1954 can be attributed to its technological innovation, national significance, versatile applications, and its role in advancing computer industry and computer science education in Japan.
Supporting texts and citations to establish the dates, location, and importance of the achievement: Minimum of five (5), but as many as needed to support the milestone, such as patents, contemporary newspaper articles, journal articles, or chapters in scholarly books. 'Scholarly' is defined as peer-reviewed, with references, and published. You must supply the texts or excerpts themselves, not just the references. At least one of the references must be from a scholarly book or journal article. All supporting materials must be in English, or accompanied by an English translation.
Bibliography
References
Note: In many articles from the 1950s, Hz (Hertz) for the unit of frequency is NOT used. Instead C or c (cycle) is used. This may require caution when one reads the old document.
[1] Hidetosi Takahasi, Eiichi Goto, Hiroshi Yamada: "On Mechanical and Electronic Calculation Methods", Technical Committee on Electronic Computers of the Institute of Electrical Communications, 1954. (In Japanese). <pr> Original Japanese title: 機械電子式計算方式について, 高橋秀俊、後藤英 一、山田博, 1954年1月29日
[Remarks] This Reference [1] is a pair with the next Reference [2]. Reference [1] is the main text, and Reference [2] is the explanatory diagram.
The explanation of this reference [1] and the following reference [2] are given in the Appendix IV.
Media:Takahashi_19540129_1.pdf
[2] Hidetosi Takahasi, Eiichi Goto, and Hiroshi Yamada: "On Mechanical and Electronic Calculation Methods (Explanatory Diagram)", The Technical Committee on Electronic Computer Research of the Society of Electrical Communicators, 1954. (In Japanese). <pr> Original Japanese title:機械電子式計算機 説明図
[Remarks] This Reference [2] is a pair with the preceding Reference [1]. Reference [1] is the main text, and Reference [2] is the explanatory diagram.
The explanation of this reference [2] and the preceding reference [1] are given in the Appendix IV.
Media:Takahashi_19540129_2.pdf
[3] Hidetosi Takahasi, Eiichi Goto: "Counting Circuits of
Parametrons", Technical Committee on Electronic Computers of the
Institute of Telecommunications, September 1955. (In Japanese).
Original Japanese title: パラメトロンの計数回路, 高橋秀俊, 後藤英一,
電子計算機研究専門委員会資料, 1955年9月26日
The following is an English summary of the content.
Abstract: Parametrons are typically designed as two-terminal elements and are primarily constructed using capacitors and transformers. These elements transition between different "energy states", with the internal state changing in response to an external driving signal (input signal). Specifically, the amplitude and phase of the input signal allow the parametron to hold two stable states (0 and 1) and perform logical operations by transitioning between them.
Counting Circuits Using Parametrons: When parametrons were used to create counting circuits, the design focused on the following key features:
(a) Counting circuits: Parametron-based counting circuits combine multiple parametron elements to implement counting.
(b) Digital logic circuits: Parametrons can serve as basic logic gates (e.g., AND, OR, NOT), which can be combined to create more complex computational circuits, such as adders and multipliers.
(c) Advantages of parametrons: Compared to conventional circuits that use vacuum tubes and transistors, parametron circuits offer enhanced stability.
[4] Hidetosi Takahasi, Eiichi Goto, Yukio Murakami, Hiroshi Yamada: "Parametron Computer PD-1516", Technical Committee on Electronic Computers of the Institute of Electrical Communications, April 1957. (In Japanese).
Original Japanese title: パラメトロン計算機, 高橋秀俊・後藤英一・村上
幸雄・山田博
電子計算機研究専門委員会資料, 1957年4月
[Remarks] PD-1516 mentioned in the paper was more like a calculator due to the initial limitation of small number of memory words. It is not a fully programmable stored-program computer of today.
Abstract: The PD-1516 is a parametron computer that was jointly developed by the University of Tokyo and Japan Electronics Instruments in October 1956. It had six words of internal storage initially. A parametron is a type of logic element that uses a ferrite core and is known for its low cost and stable operation. The PD-1516 was a calculator equipped with 16 registers, each capable of holding 15 decimal digits. The programming of this calculator was carried out using a symbolic programming language form. It contributed to the early days of computing industry in Japan.
[5] Hidetosi Takahasi, Eiichi Goto, Eiiti Wada, Takashi Soma,
Yoshihiro Ishibashi, Keisuke Nakagawa: "On the PC-1 Parametron
Computer, (1.Structure of PC-1 and 2.Program of PC-1)", Technical
Committee on Electronic Computers of the Institute of Electrical
Communications, September 25, 1958. (In Japanese).
Original Japanese title: パラメトロン計算機 PC-1 について, 高橋秀俊・
後藤英一・和田英一・相馬嵩
石橋善弘・中川 圭介 、電子計算機研究専門委員会資料, 1958年9月25日、社団法人 電気通信学会
Abstract: (Translated summary from the original Japanese by the submitter.)
The PC-1 (Parametron Computer No. 1) is a universal scientific computer that uses parametron technology, developed by the Takahasi Laboratory at the Faculty of Science, University of Tokyo. Production began in 1957 and was completed in 1958. The computer contained 4,200 parametrons, while its input/output and storage devices were sourced from various companies. Notably, the reliability of the AC magnetic core storage device is exceptional, featuring an error-correction circuit that operates normally even if a single vacuum tube fails.
The PC-1 operates internally using the binary system and supports fixed-point arithmetic. It was the fastest computer in Japan at the time (Proposer's comment: when this article published in September 1958 was written), operating 9 to 12 hours a day, with 5 hours dedicated to numerical calculations for various scientific departments, and the remaining time used for program development within the laboratory.
One of the key features of the PC-1 is its use of an AC magnetic core storage unit, ensuring stable operation. The calculation speed is enhanced by carry-lookahead circuit in the arithmetic circuit. Notably, innovations in numerical representation methods and parallel computing control have improved calculation speed by up to 2 to 3 times.
Plans for the PC-2 are already underway. Building on the experience gained with the PC-1, the PC-2 will feature enhanced memory capacity, high-speed I/O equipment, and floating-point computing capabilities. The design of the PC-1 includes a streamlined configuration of arithmetic circuits for addition, multiplication, and division, with particular attention to the addition circuit, high-speed multiplication processing, and division correction processing. The structure of the PC-1 was specifically devised to optimize efficiency and performance, and a similar approach will be applied to the design of the PC-2.
Media:PC-1_19580925.pdf
Available online at: https://www.iijlab.net/~ew/pc1/singakukai.html
The proposer believes that the above copy is a scanned and OCR'ed copy
of the original.
[6] Eiichi Goto: "Parametron Computer PC-1", Information Processing,
Vol. 16, No.1, pp. 39-43, 1975. (In Japanese).
Original Japanese title: パラメトロン計算機 PC-1 - 日本における計算機
の歴史 -, 後藤英一、情報処理, vol. 16 No. 1,
[Remarks]
Reference [6] covers the same technical content as Reference [5]
concerning the PC-1. However, this reference [6] emphasizes the PC-1's
role as a milestone in the history of computing in Japan. This
distinction arises because reference [5], written by Takahasi in 1958
during PC-1's development, contrasts with the reference [6], written
by Goto more than 15 years later, after the PC-1 had been established
as a device of historical significance.
For example,
reference [6] provides a detailed account of the development history
of the PC-1, the various challenges encountered during its creation,
and its operational methods.
The readers are cautioned that there are many typos in the English terms in this article. Media:Goto_4601_1975.pdf
[7] Hidetosi Takahasi: "Some Important Computers of Japanese Design", IEEE Annals of the History of Computing, Vol.2, No.4, pp. 330-337. Oct.-Dec. 1980, doi: 10.1109/MAHC.1980.10043.
[Remarks] Note the phrase "we were on the verge of starvation in the ruin of our defeated country", and "We were starved for knowledge as well as for food". Such was the atmosphere of post-war Japan when parametron was invented.
Abstract: Rapid growth of the computer industry is one of the most striking events in the "miraculous" industrial explosion of postwar Japan. At the time the ENIAC was completed at the Moore School of Electrical Engineering, we were on the verge of starvation in the ruin of our defeated country. When we learned of the astonishing power of the "giant brain," it seemed indeed to be something belonging to another world. We were starved for knowledge as well as for food, and some of us who were optimistic enough were inspired by this fascinating new technology. We decided to make our own computers, and the study of "mechanical brains" got under way in Japan.
[8] Chigusa Kito; "PC-1 Parametron Computer, 50th anniversary", Events and Sightings, IEEE Annals of the History of Computing, Vol. 30, No.3, pp. 74-77, July-September 2008. https://muse.jhu.edu/article/247703
[Remarks] This covers the 50th anniversary event of PC-1 that took place in 2004. Media:Event and Sightings_2008.pdf
[9] E. Goto: "The Parametron, a Digital Computing Element Which Utilizes Parametric Oscillation", Proceedings of the IRE, Volume: 47, Issue: 8, pp. 1304 - 1316, August 1959, DOI: 10.1109/JRPROC.1959.287195
[Remarks] This paper won the prestigious IRE Memorial Prize Award in Memory of Browder J. Thompson in 1961.
Abstract: The following is a brief description of the basic principles and applications of the parametron, which is a digital computer element invented by the author in 1954. A parametron element is essentially a resonant circuit with a nonlinear reactive element which oscillates at one-half the driving frequency. The oscillation is used to represent a binary digit by the choice between two stationary phases π radians apart. The basic principle of logical circuits using the parametron is explained, and research on and applications of parametrons in Japan are described.
[10] E. Goto; K. Murata; K. Nakazawa; K. Nakagawa; T. Moto-Oka; Y. Matsuoka; "Esaki Diode High-Speed Logical Circuits",
IRE Transactions on Electronic Computers, Volume: EC-9, Issue: 1, pp. 25 - 29, March 1960
DOI: 10.1109/TEC.1960.5221600
Abstract: Logical circuits using Esaki diodes, and which are based on a principle similar to parametron (subharmonic oscillator element) circuits, are described. Two diodes are used in series to form a basic element called a twin, and a binary digit is represented by the polarity of the potential induced at the middle point of the twin, which is controlled by the majority of input signals applied to the middle point. Unilateral transmission of information in circuits consisting of cascaded twins is achieved by dividing the twins into three groups and by energizing each group one after another in a cyclic manner. Experimental results with the clock frequency as high as 30 mc are reported. Also, a delay-line dynamic memory and a nondestructive memory in matrix form are discussed.
Submitter's note: "30 mc" in the abstract means "30 MHz" in today's parlance, that is, "c" stands for "cycle" Media: Goto_EsakiDiode.pdf
[11] Eiiti Wada: "The Initial Input Routine of the Parametron Computer PC-1", pp. 435-452 in Raúl Rojas, Ulf Hashagen ed: "The First Computers: History and Architectures", MIT Press, 2002.
Abstract: Forty years ago, the PC-1, parametron computer 1, was born at Professor Hidetosi Takahasi's Laboratory. The logical elements of the PC-1 were parametrons, which supported majority logic. The memory system operated in a two frequency read/write scheme. The word selection mechanism applied error correcting code to decrease the number of elements. Most of the hardware technologies were created by Eiichi Goto. We studied the EDSAC computer precisely, however we developed our own architecture and programming system based upon our own philosophy. The machine instruction set was chosen to ease programming. The normal teletype on the market was employed, leaving the burden of code conversion tasks to software, which seemed to us to have had almost infinite abilities. However, the real memory capacity was indeed very small, which forced us to invent a clever way to implement things. In this paper, after introducing the functions of the initial input routine R0, examples of (i) code conversion table parasitic on the program body and (ii) the magic number method to control the number of multiplications, both used in the initial input routine, are described. The PC-1 is one of the first computers which implemented interruption. That is, the peripheral devices would interrupt the running program by saving the address of the next instruction to be executed and jumping to a fixed location in the memory. As a simple experiment of multiple programming, cooperation of the binary to decimal conversion program and the printer control program by means of the circular buffer was performed. At the end of this paper, the program lists of the selected routines are appended.
[12] M. V. Wilkes, D. J. Wheeler and S. Gill: The Preparation of Programs for an Electronic Digital Computer. Addison-Wesley Press, Inc. 1951.
[13] Hidetosi Takahasi, "電子計算機の誕生" (Birth of an Electronic Computer in Japanese), Chuo-Koron Sha, 1971.
[14] Committee for the 50th Anniversary of PC-1, "パラメトロン計算機 PC-1 1958-2008" (Parametron Computer PC-1 1958-2008 in Japanese), https://www.iijlab.net/~ew/pc1/pc150th.pdf),
[15] Japan's first electronic computer FUJIC, Bunji Okazaki, Information
Processing Society of Japan, Vol 15, No. 8, August 1974.
Original Japanese TITLE: "わが国初めての電子計算機 FUJIC"
Available online at: https://museum.ipsj.or.jp/guide/pdf/magazine/IPSJ-MGN150808.pdf
[Remark] This was the first Japanese electronic computer using vacuum tubes.
[16] On the Application of Parametrically Excited Nonlinear Resonators
(in Japanese), Eiichi Goto,
Transaction of IEICE, Vol.38(1955), No.10, p.770-775.
Original Japanese title: 非線形共振子のパラメータ励振とその応用
[Note] This summarizes the knowledge of parametron at the early stage than in Goto's seminal paper [9]. The paper was published in 1955 after the invention of Parametron in late 1954. This paper was written before the parametron computer PC-1 was ever designed and began construction in 1957 and thus lacked any reference to complex logic circuit of computers. It seems that Goto was not aware of the similar proposal done by von Neumann using the variable capacitor back when he wrote [16], but do note that Goto already considered changing C of LC circuit to create a parametric oscillation circuit on top of L.
The paper refers to the majority logic operation of parametron, how to represent this in a diagram, and three phase excitation circuit explained in Appendix VI. So it can certainly be called a precursor of [9]. It is genuinely only about parametron logic element while [9] discusses the PC-1 computer as well. The paper lists a few advantage of parametron, namely, (1) long lifetime, (2) relatively fast operation speed, (3) low price, (4) relatively low power consumption in comparison with vacuum tubes, but, with almost clairvoyance, it also refers to potential difficulty of creating high frequency external exciton power circuit, which became the bottle neck of wide adoption of faster parametron computers (again discussed in Appendix V).
This paper does not show the detailed analysis of Mathieu's equation as shown in [9]. It claims that the detailed analysis would be given in another paper.
Abstract: Proposer's translation of the original abstract in Japanese.
We conducted theoretical and experimental studies on the parametric excitation of a nonlinear resonator using a nonlinear reactor composed of ferromagnetic or ferroelectric materials as the tuning element. As a result, we obtained a new circuit element with three functions: amplification, memory, and shaping. The author has named this new circuit element “Parametron” (parametric excitation resonator). Parametron is suitable for logic circuits and can be used to construct any complex logic devices. This paper explains the basic principles of operation and usage of Parametron. cf. Goto used ferromagnetic material to change inductance (L) of LC circuit and investigated further. If he used ferroelectric material to change capacitance (C), he would have created a variable capacitance parametron which would be proposed by von Neumann, but he did not pursue this route much further.
The original Japanese abstract:
磁性体あるいは強誘電体等より成る, 非線形リアクタを同調素子とする非
線形共振子のパラメータ励振について理論的並びに実験的に研究した結果,
増幅,記憶,整形の3作用を有する新しい回路素子が得られた。筆者はこの
新回路素子にパラメトロン (Parametron, 補助変数励振共振子)なる名称
を与えた。パラメトロンは論理演算回路に適し、如何に複雑な機能を有す
る論理演算装置でもパラメトロンのみで構成する事が出来る。この論文で
は,パラメトロンの動作原理とその使用法の基礎的事項に関して説明する。
Patents
[P1] EIICHI GOTO: "RESONATOR CIRCUITS", US Patent 2,948,818, Filing: May 16, 1955. Patented: Aug. 9, 1960.
[P2] EIICHI GOTO: "DEVICE COMPRISING PARAMETRICALLY EXCITED RESONATORS", US Patent 2,948,819, Filing: Feb. 27, 1956. Patented: Aug. 9, 1960.
Awards
[A1] IRE Memorial Prize Award in Memory of Browder J. Thompson, in 1961.
Citation: For his paper entitled “The Parametron, a Digital Computing Element which Utilizes Parametric Oscillation,” which appeared in August 1959 issue of Proceedings of the IRE. (reference [9] of this proposal document.)
The IEEE Browder J. Thompson Memorial Prize Paper Award was established in 1945 and is presented for the most outstanding paper in any IEEE publication issued between 1 January and 31 December of the preceding year by an author or joint authors under thirty years of age at the time the original manuscript was submitted. The award has been superseded by Leon K. Kirchmayer Prize Paper Award since 1997. (https://ethw.org/IEEE_Browder_J._Thompson_Memorial_Prize_Paper_Award)
[A2] Winner of the Asahi Prize (1959)
This was given for his contribution to PC-1 computer.
The Asahi Prize (朝日賞, Asahi Shō), established in 1929, is an
award presented by the Japanese newspaper Asahi Shimbun and Asahi
Shimbun Foundation to honor individuals and groups that have made
outstanding accomplishments in the fields of arts and academics
and have greatly contributed to the development and progress of
Japanese culture and society at large. (from Wikipedia,
https://en.wikipedia.org/wiki/Asahi_Prize)
Appendix I: Detailed list of commercial Parametron computers
Many commercial parties and research institutes set out to develop computers based on parametron in Japan after Goto's invention was announced. Here is the list of such computers and the current status of exhibition display, if available.
Strictly speaking communication devices are not computers, but KDDI's ARQ and others are mentioned at the end, too.
Japan Electronics Instruments
PD-1516 (1956): This was more like a calculator because it had
initially only 6 words of memory. Development division was later moved to Fujitsu.
https://museum.ipsj.or.jp/en/computer/dawn/0051.html
The University of Tokyo (Takahasi Laboratory)
PC-1/4 (1957): A preliminary experimental model of the PC-1, about the size of a notebook. It can perform arithmetic on binary numbers each with 9 bits. The input device has only 7 toggle switches. PC-1 was going to use a binary number of 36-bit. This machine used 9-bit number, thus the name, 1/4.
PC-1 (1958): A full 36-bit stored-program machine with a 2 m wide and 1.5 m high chassis. ([6]). A female employee responsible for relay wiring at Fuji Tsushinki (today's Fujitsu) produced 4,300 parametrons. The excitation frequency is 2 MHz, and the operating frequency is 15 kilohertz. The storage device has 256 bytes. The instruction set includes simple addition, subtraction, multiplication, and division. It features a "high-speed carry lookahead circuit" for fast addition, multiplication, and division, based on Goto's own idea, and "proactive control" for performing multiple instructions simultaneously. Note that the "high-speed carry lookahead circuit" was being studied all over the world at then, and Goto et al used a different terminology then. The input and output use 6-hole perforated tape. The first program is an "neglect erase" that skips only the parts where no data is written (which was represented by all holes in 6 positions, known as erase code) and copy the rest to a new tape. (Eiiti Wada, page 1, [14]) Many researcher from outside the University of Tokyo came to use PC-1, and it was used for programming training, too. It has been dismantled and no longer exists.
PC-2 (1960): Funded by the Ministry of Education, it was jointly developed with Fujitsu. It is an enhanced version of the PC-1 for scientific calculations, using 13,000 parametrons. The excitation frequency is 6 MHz, and the operating frequency is 60 KHz. The word length is 48 bits. It includes functions such as floating-point arithmetic, data search, and multiplication unit that uses multiplication of 4 bits by 4 bits entity. It was the fastest parametron computer and outperformed the transistor-based ETL Mark IV A (although junction transistors were still at a speed disadvantage at the time) [7]. It took 9 seconds for a 1000-digit calculation of the base of the natural logarithm, e. It was commercialized as FACOM 202. FACOM 202 had four times as many memory words as PC-2, and performed two to three times as fast as PC-2 when fixed point addition and multiplication were performed.
https://museum.ipsj.or.jp/en/computer/dawn/0035.html
There is a short description of how FACOM 202 was used at the Institute of Solid State Physics in the newsletter of the Institute, p. 17, No. 5, Vol 4, December 1964, in Japanese, available online at https://www.issp.u-tokyo.ac.jp/maincontents/docs/tayori/tayori04-5.pdf. A notable feature of FACOM 202 was the use of Algol language compiler which made it easier for researchers to use it than the machine language assembler.
Nippon Telegraph and Telephone Corporation (Telecommunications Research Institute)
MUSASINO-1 (1957): Initially had only 32 words using magnetic core memory (later expanded to 256 words). The instruction set was based on ILLIAC 1.
Despite the meager 32 words initially, MUSASINO-1 did show that
parametron can be used to build a complex circuit such as CPU and it
also acted as a test bed for parametron's durability. Later, in 1960,
NTT produced a commercial version called MUSASINO-1B that required
much less maintenance, which was later sold as FACOM 201 by today's
FUJITSU. NTT produced a billing machine using parametron later, too.
https://museum.ipsj.or.jp/en/computer/dawn/0013.html
MUSASINO-1B (1960): Jointly developed with Fujitsu. Commercialized as the FACOM 201.
CAMA (1963): For call billing only, not programmable.
https://museum.ipsj.or.jp/en/computer/dawn/0032.html
Hitachi
HIPAC MK-1 (December 1957): The company's first computer.
https://museum.ipsj.or.jp/en/computer/dawn/0015.html
A model is displayed at the Central Research Laboratory of Hitachi
Corporation as of Feb 2025.
HIPAC 101 (1960): Commercialized.
https://museum.ipsj.or.jp/en/computer/dawn/0019.html
HIPAC 103 (August 1961): Commercialized. For scientific and technical calculations.
https://museum.ipsj.or.jp/en/computer/dawn/0041.html
Numerical Control Unit of milling machine
Hitachi also did research on numerically controlled machining tools. Only the listing of relevant articles from that time is given below. The submitter feels that by the time the software, sensors and actuators were in place, parametron computer was not so attractive to be the embedded computer for numerically controlled machine. That is why the last article quoted below describes transistor-based control units while the first and second articles discussed parametron-based numerical control unit. This shift reflects the changing device technology very well.
Review of parametron written in June 1958 by a researcher at Agency of Industrial Science and Technology (in Japanese): https://www.jstage.jst.go.jp/article/sicejl1954/4/1/4_1_39/_pdf/-char/ja
This review appeared only 4 years after parametron was invented in 1954
and after numerical control unit(s) were designed and was written from
the parametron user's point of view in 1958.
In this article, a control unit consisting of 900 parametrons is
described.
Article written in 1958 by Hitachi researchers:
Numerical Control of Machine Tools, Masasuke Ogawa and Yoshiro Anno, September 1958, special issue no
25. of Hitachi Review (日立評論 in Japanese) on Machining tools
https://www.hitachihyoron.com/jp/pdf/1958/ex25/1958_ex_25_04.pdf
Original Japanese title: 工作機械の数値制御
Abstract
米国,英国における数値制御工作機械の最近の進歩について概論し,数値制御方
式を工作機械に適用する場合の技術的問題点をあげた｡特に機械技術者の立場よりプログラミングとパルス分配の問題,ア
ナログ方式とディジタル方式の得失,パワーサーボ機構の選定上の問題点を明らかにした｡あわせて工
業技術院機械試験所が試作研究し,日立製作所の協力した数値制御竪フライス
盤 (milling machine) の内容を紹介した｡
The above article in September 1958 describes
a numerical control unit for a milling machine built at Agency of
Industrial Science and Technology using parametron.
However, in the article written in 1960 quoted next, Hitachi revealed
a control unit made of transistors. The technology was changing
quickly around 1960.
An article written in 1960 both by a researcher from the government,
Agency of Industrial Science and Technology and people at Hitachi's
factories on the lately finished HIDAM-402, a numerical control unit
(in Japanese). This unit no longer uses parametron(!) :
https://www.hitachihyoron.com/jp/pdf/1960/ex34/1960_ex_34_04.pdf
NEC
NEAC-1101 (1958): The company's first computer.
https://museum.ipsj.or.jp/en/computer/dawn/0017.html
NEAC-1102 (1958): Jointly developed with Tohoku University and
delivered to the Electrocommunications Research Institute, Tohoku
University. Also known as SENAC.
It was delivered to Tohoku University in March 1958, but was not made
available for full scale operation until November 1958.
https://museum.ipsj.or.jp/en/computer/dawn/0020.html
NEAC-1103 (1960): Delivered to the National Defense Agency Technical Research Laboratory.
NEAC-1201 (1961): Commercialized as an office computer. Its successors
were NEAC-1202 and NEAC-1210.
NEAC-1201 was quite a success as a small computer at that
time. It may be better called tabulator because it had only 120 words
although it WAS a stored program computer.
More than 800 units were sold which was beyond the
initial expectation of 200-300 units sold in total according to "History
of First and Second Generation Japanese Computers and the Preservation
of (Early) Examples" by Akihiko Yamada, in "技術の系統化調査報告 第１
集" （"Systematic Study of Technology Report", Volume One in Japanese)
by National Museum, March 2001.
The successor 1210 sold well, too.
https://museum.ipsj.or.jp/computer/dawn/0040.html (in Japanese)
NEAC-1210 (1964): It sold well. 700+ units sold by August 1966.
https://museum.ipsj.or.jp/computer/dawn/0060.html (in Japanese)
Oki Electric
OPC-1 (1959):
https://museum.ipsj.or.jp/en/computer/dawn/0023.html
INS-1 (circa 1962): Installed at Japan Nuclear Research Institute
(which was established in July 1955).
The sketchy description of manufacturing and installation of this
computer is only found in the following Japanese page.
https://www.hpcwire.jp/archives/40591 (in Japanese)
Fujitsu
FACOM 200 (September 1958): A prototype that of computers that followed.
https://museum.ipsj.or.jp/en/computer/dawn/0062.html
FACOM 212 (Shipped in April 1959): Commercialized as an office computer. One word was 12 decimal digits. 32 words of magnetic core memory. One word could hold 4 instructions. So, in total, 128 instructions could be used.
FACOM 201 (1960): Commercialization of MUSASINO-1B.
A machine is on permanent display in 2018 at Tokyo University of Science.
A photo is in the following URL.
https://www.jaima.or.jp/resource/jp/heritage/pdf/2016_No66_67.pdf
Quote of the paragraph in the URL.
The parametron, invented in Japan as an original, unique logic element, was used in a number of Japanese computers as a cheap and robust logic unit in the transition period from the vacuum-tube era to the transistor era. It is of tremendous value as technological heritage. Approximately 6,000 parametron units are used in this computer.
This computer was utilized in Tokyo University of Science for the purposes not only of scientific computations but also of technology developments such as “effects of the shape of a small rocket on the propulsion by solid fuel combustion” and “theoretical computations of the strength of car-body structures”. Furthermore, it was employed for the education of computer technology and contributed to training engineers.
Display of a large computer: FACOM 201
https://www.tus.ac.jp/info/setubi/naruhodo/main/calculators.html
(in Japanese)
Tokyo University of Science even organized an event on "Parametron
computers and relay computers." in 2018.
https://www.tus.ac.jp/info/setubi/museum/event_data/2018parametron/2018parametron.html
(In Japanese)
[Remarks] The above event reflected the impact parametron had on the Japanese
computer industry and education in the late 1950 and the 1960s.
FACOM 202 (1960): Commercialization of the PC-2. For scientific and technical calculations. At the time of completion, it became the fastest computer in Japan.
Fujitsu also built numerical control unit for machining tools in 1956. That division later became today's FANUC. The control unit built in 1956 used parametrons. But it was a demonstration prototype. However, later models that were deployed in the field did not use parametrons.
The following is the history page of today's FANUC.:
https://www.fanuc.co.jp/en/profile/history/
The entry for 1956 reads "The first NC and SERVO systems in the
Japanese private sector were developed successfully."
This NC used parametrons.
The entry for 1958 reads "The first commercial FANUC NC was shipped to Makino Milling Machine Co., Ltd." The NC used vacuum tubes and electromechanical relays. (page 3 of Japanese PDF at https://www.jspe.or.jp/wp/wp-content/uploads/activity/int05_2.pdf, Technology and Management, Seiemon Inaba, the president of FANUC)
The technology was changing rapidly in late 1950s and early 1960s.
Mitsubishi Electric
MELCOM 3409 (1960): Mitsubishi immediately switched to using transistor after this machine, and not much material is available online.
Aside from the computer business, however, Mitsubishi developed ARQ (Automatic Repeat-reQuest) for telex operation for KDD (precursor of today's KDDI). KDD's ARQ was the world's first electronic ARQ and became popular and as such was exported to other countries.
For example, Mitsubishi Electric delivered a parametron ARQ to KDD, the predecessor of today's KDDI in September 1958 according to https://shashi.shibusawa.or.jp/details_nenpyo.php?sid=5820&query=&class=&d=all&page=45
Mitsubishi Electric continued elaborating parametron ARQ devices later. A report written in June 1963 about a parametron ARQ used in Japan at the time is Available online at: https://www.giho.mitsubishielectric.co.jp/giho/pdf/1963/6306.pdf
Its abstract in English (original):
Data transmission Without error has come to the front through the centralization of business transaction. The data transmission equipment has come to play an important role of the IDP system. Type TZ-11 1ow speed data transmission equipment is the expansion of the ARQ system extensively used for international communication. By correcting errors on the transmission circuit, and being provided with function of correcting errors including those of input and output devices from the reading to the punching of paper tape, they constitute ideal transmission circuits. This equipment is installed corresponding to a 6-unit printing telegraph circuit, and provided with PARAMETRONS (emphasis by the submitter) at its principal part, it has secured for better reliability and very stable operation for long, continuous operation.
This Mitsubishi report in June 1963 refers to an earlier report of
KDD's ARQ device, which the submitter of the application believes was
the exported version. But unfortunately the submitter could not find
the original KDD's research lab's report in time for final submission
and editing.)
Koden Electronics Co., Ltd.
KODIC-401 (1960): Experimental prototype.
KODIC-402 (1961): General-purpose computer. Commercialized with an operating
frequency of 2 MHz, decimal 16-digit fixed-point stored programming,
and a magnetic drum storage device with 4000 words. A total of three
systems were delivered for in-house use, including one to the Faculty
of Engineering at Japan University and another to the Department of Industrial
Engineering, Faculty of Engineering, Osaka Electro-Communication
University (OTSUDAC-1, delivered in March 1963, currently on display
and preserved).
OTSUDAC-1 was on display at the university in 2014. (https://www.hpcwire.jp/archives/41516)
The following is a press release (in Japanese) regarding the chair of Koden Electronics, Mr. Itoh,
visiting the display on July 14, 2014.
https://www.osakac.ac.jp/news/2014/327
[Remarks] The following excerpt showed the impact of parametron-based computer on education at universities in Japan circa 1960.
An excerpt from the release (translated by the proposer):
... This electronic computer was produced by Koden Co., Ltd. around 1963, and is said to be very valuable as an early computer using a parametron element at a time when computers were not yet widespread. The University was the first university to introduce this computer for education and research, and has preserved it on the second floor of Building M at the Neyagawa Campus since 1976 to commemorate the achievements of that time. ...
KDDI (formerly KDD)
KDDI was one of the early backers of parametron after Goto's first presentation of parametron. Thus, it developed many interesting devices that used parametron.
KDDI developed Morse code to Teletype code converter using parametron in time for Melbourne Olympic in 1956. It used the converter to transmit news messages from Olympics. The machine(s) worked flawlessly.
Encouraged by the success of the device, KDDI developed various devices using parametrons. Most famous are the ARQ (Automatic Response Request) device for telex operation and regenerative signal repeater used for long distance transmission.
Parametron-based Regenerative Relay machine
Let us first explain about the regenerative signal repeater since not much information is available online about it.
KDDI's website had a snippet from the old KDD company history, and in it, the description of regenerative relay machine. https://www.kddi-research.jp/sites/default/files/labo/history/1957a_%E3%83%91%E3%83%A9%E3%83%A1%E3%83%88%E3%83%AD%E3%83%B3%E5%BF%9C%E7%94%A8%E6%A9%9F%E5%99%A8%E3%81%AE%E5%AE%9F%E7%94%A8%E5%8C%96.pdf Submitter's translation of the above short document.: Practical application of parametron-based devices To prevent the increase in distortion that occurs during multi-stage relaying of telegraph signals, it was necessary to regenerate and relay the signals at relay points. Although electronic tube-based devices had already been put into practical use, the invention of the Parametron led the research into applying it to regenerative repeaters. In 1956 (Showa 31 in Japanese calendar), prototype testing and field experiments were conducted, resulting in the development of a regenerative repeater incorporating a new idea that automatically tracks variations in telecommunication distortion and significantly increases the reception margin. This device was deployed in the field in 1957. Source: KDD Company History
Parametron-based ARQ
The following Japanese URL suggests the date when the first parametron ARQ device was completed at KDD as April 1956. : https://shashi.shibusawa.or.jp/details_nenpyo.php?sid=13210&query=&class=&d=all&page=7 This was only two years after the invention of parametron in 1954.
Also, Mitsubishi Electric developed ARQ devices for KDD. For example, Mitsubishi delivered a parametron ARQ to KDD, the predecessor of today's KDDI in September 1958 according to https://shashi.shibusawa.or.jp/details_nenpyo.php?sid=5820&query=&class=&d=all&page=45
KDDI's website had a snippet from the old KDD company history.
Here is one about Parametron-based ARQ.
https://www.kddi-research.jp/sites/default/files/labo/history/1958a_%E3%83%91%E3%83%A9%E3%83%A1%E3%83%88%E3%83%AD%E3%83%B3%E5%BC%8FARQ%E8%A3%85%E7%BD%AE.pdf
Submitter's translation of above document.:
Parametron-based ARQ device
Since its inception, KDD has introduced ARQ (automatic error correction) into telex and dedicated telegraph lines, promoted the domestic production of ARQ devices, and completed a fully electronic tube-based ARQ in 1956 (Showa 31 in Japanese calendar). Further research was conducted to achieve stability and miniaturization, leading to the development and practical application of a Parametron-type ARQ device using Parametron elements (logic circuit elements invented by Dr. Goto of the University of Tokyo in 1954) in 1958. This ARQ device incorporated error correction function and could be used in dual time-division multiplex fashion, and by combining two such devices, it could also be used in quad time-division multiplex system. Compared to the conventional TZ-2 type electronic tube-based ARQ device, it achieved extremely high performance with power consumption reduced to 1/4 and floor space reduced to 1/3.
Source: KDD Company History
The above short document explains somewhat different dates from the earlier description, but we can see that very advanced ARQ was created using parametron in the late 1950s by KDDI.
It is not clear until when the parametron-based devices were used. Some seem to have been used in the mid and late 1970s (anecdotal oral communication.)
Luckily, there *IS AN INDIRECT WRITTEN RECORD* of JICA, an government-backed agency to offer assistance to developing countries at the Japanese government website. Its record of circa 1967 - 1968 shows that KDDI offered technology seminars regarding long distance communication using Telex and its management. and in that seminar, parametron is clearly mentioned along with ARQ. Thus KDDI used parametron-based ARQ and other devices still in 1967 and possibly years beyond that. URL of JICA document (in Japanese): https://openjicareport.jica.go.jp/pdf/10015782_02.pdf
A summary of a seminar topic in the above JICA document is as follows. (Original Japanese was translated into English.)
Under the larger heading of "Electric communication", there are "Telex", and "Short wave radio" entries.
Telex: The purpose of the course and training content.
The communication technology employed for Telex in Japan is introduced to contribute to the communication technology of the seminar attendees' countries. The training is done by the classroom lectures on electric communication systems, PARAMETRON (EMPHASIS by the author), ARQ facilities, and hands on training for network path monitoring, frequency modulation, malfunction repair, ARQ facility maintenance, etc.
And, believe it or not, in 1977, a similar seminar that includes
"parametron" in its topic was still offered by KDDI
via government backing (!).
The following URL is at the Ministry of Internal Affairs,
Communications of Japan.
https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/s53/html/s53a02080202.html
It lists the white paper on communications (in Japanese) of the year 1978, which means it covers the activity up to 1977 including 1977. In its chapter 8, section 2.2, it is mentioned that there was a seminar extending from August 18, 1977 to November 20, 1977, given 16 times on "International Telex communication technology". The description of the seminar states that the objective of the seminar is to train the attendees to acquire necessary technical knowledge and skill on PARAMETRON (emphasis by the submitter), transistor, integrated circuit, ARQ device, and Telex device in general. It was offered to 12 people from 11 countries.
The seminar in the second half of 1977 means the parametron-based ARQ
was still used in full swing at KDDI in 1977.
The ARQ had a very long life time of more than 20 years (!).
The white paper of next year (1979) no longer mentions "parametron".:
https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/s54/index.html
It seems that the device was decommissioned in 1978 or the importance
of Telex communication itself became lighter in view of other
communication protocols defined by CCITT. 1977 and 1978 saw the surge
of early personal computers such as Apple-I, TRS-80, and others.
Thus the communication based on RS-232C
and acoustic modems became very popular in certain areas, and its
speed was comparable to telex or faster.
The communication landscape was changing.
KDDI ceased its Telex service in March 2005.
Wire Memory
Notable development done by KDDI is the development wire memory.
KDDI's website had a snippet from the old KDD company history about this.
https://www.kddi-research.jp/sites/default/files/labo/history/1962c_%E3%83%AF%E3%82%A4%E3%83%A4%E3%83%91%E3%83%A9%E3%83%A1%E3%83%88%E3%83%AD%E3%83%B3.pdf
Translation of the above short document by the submitter.
Wire Parametron
In 1962 (Showa 37 in Japanese calendar), the wire parametron and thin-film parametron were developed. Research on magnetic thin films using vacuum deposition began around 1956, and in 1957, a parametron was constructed using cylindrical Permalloy thin films, achieving successful oscillation. Additionally, we developed a method for producing magnetic thin films using the electroplating process, and in 1962, we successfully achieved oscillation of a wire parametron by plating magnetic thin films onto fine copper wires. The following year, in 1963, we developed wire memory using this wire parametron and commercialized it as high-speed memory for computers.
Source: KDD Company History
KDDI also has a description of a device that followed up on the wire
memory.
https://www.kddi-research.jp/sites/default/files/labo/history/1969c_%E3%83%95%E3%82%A1%E3%82%A4%E3%83%B3%E3%82%B9%E3%83%88%E3%83%A9%E3%82%A4%E3%83%97%E3%83%A1%E3%83%A2%E3%83%AA.pdf
Permalloy Electroplated Wire Memory Matrix and Fine Stripe Memory
In 1963 (Showa 38 in Japanese calendar), we developed wire memory using a base copper wire coated with electroplated Permalloy. Previously, memory using ferrite cores had been used, but memory utilizing magnetic thin films such as Permalloy offered the significant advantages of high-speed operation and compact, lightweight design. Furthermore, in 1969, we developed fine-stripe memory that was extremely compact and operated at high speed by leveraging vacuum deposition technology, electroplating technology, and micro-fabrication technology. These magnetic thin film technologies later led to applications in optical magnetic memory using amorphous materials.
Source: KDD Company History
Thus, KDDI as an early backer of parametron did interesting research and development in 1960s using parametron at central theme.
Appendix II: subroutines and the type of calculations done on PC-1
These are taken from references [6] and [14].
Takahasi/Goto Lab:
Interrupt handling multitasking
Fast Fourier Transform (very similar to the FFT known later)
Elliptic function table based on summation rule
Exact calculation of integer arithmetic of arbitrary length using modular arithmetic
Numerical simulation of Goto pair
Created by others:
Eigenvalue solver for symmetric and asymmetric matrices
(Solving 10x10 matrix problem in 512 words PC-1!)
Partial differential equation solver
Nuclear (Magnetic) Moments of atoms
Calculation in two coulomb center potential field
Dispersion Relations in Nucleon scattering
Crystal structure analysis
Electron beam diffraction analysis of gas
Molecular vibration analysis for spectroscopy
Among the researchers who did the molecular vibration analysis
was Mitsuo Tasumi (1937 – 2021).
He received the prestigious Optics Society of America's Ellis
R. Lippincott Award in 1999“for outstanding contributions to
vibrational spectroscopy in studying the structures and
dynamics of synthetic polymers, proteins, photosynthetic
systems, and a number of related small molecules.”
https://www.optica.org/History/Biographies/bios/Mitsuo_Tasumi
His thesis calculations were done on PC-1 and PC-2.
He left his memoir regarding the 50th anniversary of PC-1 in
2008, and stated that PC-1 has determined his career.
http://sapiarc.web.fc2.com/Essay/2008/2008-03.pdf (in Japanese)
There were many more, but Goto stated in [6] (written in 1975) that many researchers who did the calculation on PC-1 were scattered around the world and he could not obtain much information already in 1975. He specifically mentioned detailed information on ordinary differential equation solver, and solver for partial differential equation for magnet design were missing from [6].
Appendix III Parametrons in Disguise
Original parametron invented in 1954 used ferrite core. The idea of parametron, i.e., parametric oscillation, recurred later from time to time in different forms in designing digital circuits and certain applications.
Goto Pair
When Tunnel diode (https://en.wikipedia.org/wiki/Tunnel_diodewas) was invented, Goto created a new device consisting of two tunnel diodes called Goto pair. (See the reference [10], E. Goto; K. Murata; K. Nakazawa; K. Nakagawa; T. Moto-Oka; Y. Matsuoka; "Esaki Diode High-Speed Logical Circuits")
This is actually a parametric device of a sort. The logic device achieved 30 MHz operation which was quite an impressive achievement back in 1960.
Quantum flux parametron
Much later, Goto realized a parametric oscillation can be observed in a super-cooled Josephson-junction device, and devised a logic device called quantum flux parametron in 1986, and carried out research on it. (https://en.wikipedia.org/wiki/Quantum_flux_parametron)
This was explained in the "Features that set this apart from other similar achievements" section, and the version of QFP called Adiabatic Quantum Flux Parametron (AQFP), with its energy-saving feature, seems to attract attention again in energy-aware computing industry of today. [REF: Adiabatic Quantum-Flux-Parametron: A Tutorial Review, https://www.jstage.jst.go.jp/article/transele/E105.C/6/E105.C_2021SEP0003/_article ]
These and other contributions by Eiichi Goto to the development of computer technology are chronicled succinctly in a web page of Japan Information Processing Society after he passed away in 2005. Interested readers are referred to the following URL.: https://museum.ipsj.or.jp/en/pioneer/gotou.html
Appendix IV: Earlier research for building computer at Takahasi laboratory
The team at Takahasi laboratory including Eiichi Goto was very interested in building a computer of their own after their study of EDSAC and other computers abroad. Articles written by them showed their interest in novel, reliable and low cost logic elements which would let them build a computer at the University of Tokyo, and once the parametron was invented and its reliability and ease of maintenance far exceeded both the vacuum tubes and transistors, their interest turned into building a computer using parametrons.
For example, the papers by Hidetosi Takahasi, Eiichi Goto, and Hiroshi Yamada in 1954 [1] (Hidetosi Takahasi, Eiichi Goto, Hiroshi Yamada: "On Mechanical and Electronic Calculation Methods", Technical Committee on Electronic Computers of the Institute of Electrical Communications, 1954.) and [2] (the explanatory diagram that accompanied [1]) BEFORE the invention of parametron described the approach regarding information storage that combined mechanical and electronic methods.
The paper [1] and its diagram [2] describe a "Quantum Voltage Storage Device".
The term "Quantum" is used with the meaning of "discretized" or "discrete", or "digital" in today's parlance. The authors' background in physics shows here. Takahasi was a professor at the department of physics and thus "quantum" was a familiar term to the authors.
Readers in the 21st century (in 2025) may find the description in [1] and [2] very banal. It is common knowledge today. But back in 1954, it was an important overview of information storage via electronic and legacy mechanical element (read mechanical rotating switch of the time).
With previous research like this, the laboratory members began working on building components of computers using parametron once parametron was invented and its reliability was proven.
The following is the translation of the first paragraph by the
proposer.
Please recall this was in 1954 and electromechanical relay computers
were still in vogue. They actually tried to use a mechanical switch
from a university PBX for the mechanical scanning device that scans
the storage devices and restores the electric charge before the
electric charge is lost and the information is lost. Basically, they
described an electromechanical equivalent of periodic refreshing of
dynamic RAM in today's terminology.
Also, please note that the office automation would be far into the future. The paper was written manually. All the characters were hand drawn and mimeograph was used to print the paper for distribution. (Japan did not have a very efficient and compact typewriter for its characters as in English-speaking countries. Until the early 1980s when PC finally swept the office scene, most of the papers were hand-written although there WERE so-called Japanese typewrite which is much more cumbersome than the western equivalent due to the large number of characters used in Japan.)
1. Quantum Voltage Storage Device". Since the electric charge stored in a storage device hardly changes in a short period of time, it is clear that information can be stored for a short period of time. However, if left unattended for an extended period of time, the electric charge leaks out and the information is lost. To prevent this, the charging voltage of the storage device should be quantized to a discrete value, and before the information is lost, the electric voltage of the storage device should be restored to the reference discrete value by an appropriate device. A machine that has the function of restoring the charging voltage of the storage device to the standard discrete value will be called a quantizer. It is not necessary to have one quantizer Q attached to each storage device. This is because a single quantizer can be used to quantize the voltages of many storage devices by sequentially switching between them if a suitable scanning device is available. If the time that a storage device can be left without losing information is T, and the required operating time per storage element of the quantizer and switching device S is t, then one quantizer Q can handle a maximum of N = T/t storage devices. In principle, a switching machine with such a function could be electronic or mechanical. From the above point of view, the Williams Memory Tube can be regarded as an electronic storage device. A normal Flip Flop is a storage device with a distributed capacity, but without S, and each storage device can be considered to have one quantizer Q attached to it. A Whirl Wind Tube can also be thought of as having an infinite number of Qs. However, these are all purely electronic, and there seems to have been no description yet of the second possibility above, i.e., a quantum voltage storage device that uses mechanically driven electric contact switches. Therefore, we would like to examine this new storage system and its application to computers.
And the paper continued to explain the use of mechanical element. You can see the mechanical scanning of the charge device that stores information in the diagram of [2], such as in Figure 1 of the paper.
This topic was discussed as part of the team's vision of building a
computer using available devices. This time for storage purposes.
For PC-1, though, the team used so-called two frequency AC-driven core
memory system that used non-destructive reading method, which worked
well with parametron device.
Appendix V: Speed Comparison
Here is a bit more detailed discussion on the speed than in the main text of the application.
Comparison of speeds of typical Japanese computers around the time
parametron was invented (1954) and PC-1 started operation (1958) is
shown.
The upper part is quoted verbatim from Goto's seminal paper [9] ("The
Parametron, a Digital Computing Element Which Utilizes Parametric
Oscillation", Proceedings of the IRE, published in 1959). It lists the
speed of typical parametron-based computers of that time.
The lower part of the table is created using data taken from a table, Table 2.2 in the report of National Science Museum: "History of First and Second Generation Japanese Computers and the Preservation of (Early) Examples", Akihiko Yamada. (available online at: https://sts.kahaku.go.jp/diversity/document/system/pdf/003.pdf)
Lower part is created to mimic the original Goto's table.
Unfortunately, some numbers in the National Science Museum report did not match those of Goto's table for PC-1 (the proposer trusts Goto's number), and did not match numbers for the computer built at Osaka university. Osaka University computer development is described at https://museum.ipsj.or.jp/heritage/handai-shinkukan.html (in Japanese). The addition speed of Osaka computer machine is mentioned as 40 milliseconds there. I take the number at IPSJ site as correct. There seems to be a transcription error somewhere.
To be frank, the detailed data for machines created and operated 60-70 years ago is hard to come by.
So the proposer has to be careful, but the numbers shown for FUJIC in National Science Museum report FUJIC matches the numbers in the article written by FUJIC's inventor. So, the proposer assumes they are correct.
The above was the choice of numbers in the lower part of the table.
The table shows that PC-1 (and subsequently PC-2) did well in terms of performance, but we can also tell that transistor computers with faster operation would take over once they became stable.
One computer worth mentioning in this application document for "Parametron, 1954" is FUJIC, built almost single-handedly by a developer, Bunji Okazaki, at Fuji Photo Film Co, Ltd. in July 1956 using vacuum tubes. (FUJIC is one of the only two vacuum tube computers ever built in Japan and became the FIRST electronic computer in Japan.)
The vacuum computer FUJIC was the only machine built before PC-1 that beat the speed of PC-1 in the arithmetic (addition and multiplication).
(The proposer excludes ETL Mark III from this comparison. It used
point-contact transistors, which were no longer produced due to their
unreliability, and retired as soon as Mark IV was built. The proposer
does not believe it did any meaningful calculation as opposed to PC-1
or FUJIC. It seemed to run mostly benchmarks to calibrate the machine.
There is a very sketchy description of ETL Mark III:
Transistor Computer (ETL Mark III-VI)
https://museum.ipsj.or.jp/guide/pdf/magazine/IPSJ-MGN170212.pdf
Original Japanese title: トランジスタ計算機 ETL Mark III-VI
My intention here is to compare the practically used computers for many
applications, and ETL Mark III does not fit the bill.)
Completed in July 1956, FUJIC operated as the calculating machine for optical ray tracing for lens design at Fuji Photo Film. Such calculation used to be done by female calculators working in a pair to compare the results for verification at the end of calculation steps.
Like MUSASINO-1, FUJIC was not used outside the company. Its operation was strictly inside although there were times some calculation jobs were contracted from outside parties. But the programming and operation seems to have been done only in Fuji Photo Film's Odawara works.
FUJIC's lifetime was much shorter than PC-1. Fuji Photo Film decided to stop the lens design in 1958 (or 1957? The description differ among documents), and handed over the operation to its subsidiary. However, because the operation of FUJIC was restricted to the original office, it seems that not enough technical expertise existed outside to use FUJIC. The reliability issues caused by the use of vacuum tubes made it difficult to use it, and the computer was finally donated to Waseda University in September 1958 after less than 2 and half years of operation. (Documents and records differ on how many years FUJIC operated. Okazaki mentioned that it was used two and half years [15] and so the proposer would adopt that figure. ) Okazaki, the inventor, left Fuji Photo Film in 1959, and joined NEC to pursue computer design there.
By the time PC-1 became operational in March 1958, the proposer thinks mothballing plan started for FUJIC for eventual transfer to Waseda University. That is why Takahasi wrote in September 1958 in [5]("On the PC-1 Parametron Computer"), "currently, PC-1 is the fastest computer in Japan..." (proposer's translation). FUJIC was not operational when Takahasi wrote the draft of the article because it was being transferred to Waseda University.
FUJIC was a great invention done by a person, but due to the company's decision to leave lens design, it lost backers behind its operation. It could not leave a lasting impact unlike parametron computer PC-1 used by many researchers inside and outside the University of Tokyo. Many academic papers were written about PC-1 while there is few articles that remain today about FUJIC.
References for this appendix V: Memoir by the inventor of FUJIC, Bunji Okazaki:
Japan's first electronic computer FUJIC, Bunji Okazaki, Information
Processing Society of Japan, Vol 15, No. 8, August 1974.
(original title is "わが国初めての電子計算機 FUJIC")
Available online at: https://museum.ipsj.or.jp/guide/pdf/magazine/IPSJ-MGN150808.pdf
The above documents lists the time for addition/subtraction as 0.1
milliseconds, and 1.6 milliseconds (average) for multiplication, which
match the numbers in National Science Museum report.
Bunji Okazaki at Information Processing Society:
https://museum.ipsj.or.jp/en/pioneer/okazaki.html
FUJIC at Information Processing Society page:
This page lists 0.1 msec for addition/subtraction, and 1.6 milliseconds for
multiplication. The fast speed is attributed to parallel arithmetic circuits.
https://museum.ipsj.or.jp/en/computer/dawn/0010.html
History written at the website of Fujifilm Co. Ltd (in Japanese).
日本語タイトル： カメラ・工学機器事業基盤の確立
https://www.fujifilm.co.jp/corporate/aboutus/history/ayumi/dai2-09.html
There is a description about FUJIC.:
カメラのレンズには、明るく、シャープな画像を得るために、屈折率の異 なる何枚かの単体のレンズが組み合わせて用いられる。これらの組み合わ せを決め、最適のレンズを設計するためには、複雑な計算を必要とし、高 級レンズの設計には専門家でもその計算に数か月を要するほどであった。 この計算を迅速かつ正確に行なうために、当社は、電子計算機を活用する ことを計画し、1949年（昭和24年）、その設計に着手した。試行錯誤を繰 り返しながら開発を進め、1953年（昭和28年）、組み立てを開始した。長 さ4m・高さ2mのパネルの中に、計算装置・記憶装置・制御装置が1,700本 の真空管と5,000mの配線でつながれ、使用部品は実に2万個を数えた。 1956年（昭和31年）7月に完成し、“FUJIC”と命名した。
“FUJIC”は、当社のレンズ計算に貢献するとともに、気象庁や各大学からの計算依頼にも応え、国産電子計算機の第1号として注目を浴びた。
しかし、“FUJIC”は真空管式なので、その寿命の点で実用上問題を残し ていた。このため1957年（昭和32年）、レンズ部門を富士写真光機に移設 した際、“FUJIC”を研究用として早稲田大学に寄贈した。現在は、わが 国科学史上の重要な記念として東京上野の科学博物館に展示され、同所で コンピューターのその後の発展を静かに見守っている。
English Translation of the above by the proposer.:
Camera lenses use several individual lenses with different refractive indices to produce bright, sharp images. Determining the optimal combination of these lenses and designing the best lens requires complex calculations, and it used to be that even experts needed several months to complete these calculations for high-end lenses. To perform these calculations quickly and accurately, our company planned to utilize electronic computers and began design work in 1949. After repeated trials and errors, development progressed, and assembly began in 1953. Inside a panel measuring 4 meters in length and 2 meters in height resided the computing unit, memory unit, and control unit consisting of 1,700 vacuum tubes connected by5,000 meters of wiring, with a total of 20,000 parts used. It was completed in July 1956 and named “FUJIC.”
“FUJIC” contributed to our company's lens calculations and also responded to calculation requests from the Meteorological Agency and various universities, gaining attention as the first domestically produced electronic computer.
However, as a vacuum tube-based system, “FUJIC” had practical limitations due to the limited lifespan of tubes. Therefore, in 1957, when the lens department was relocated to a subsidiary which later became Fujinon of today (after a few incarnations), “FUJIC” was donated to Waseda University for research purposes. Today, it is displayed at the Science Museum in Ueno, Tokyo, as an important historical artifact in the history of science in Japan, quietly witnessing the continued evolution of computing technology.
Note: Fuji film's description state that the lens department was relocated to its subsidiary in 1957. It did not mention the donation to Waseda University that occurred in September 1958 clearly.
Reliability issues: In contrast to PC-1's parametron stability, Okazaki mentioned that he had to replace two to three vacuum tubes each day because they got broken. (Japanese interview available online at https://ascii.jp/elem/000/001/214/1214060/2/ )
Appendix VI: Power Consumption, Excitation Frequency, Parametron Operation Speed
The historical significance of the parametron is sufficiently explained in the main text. However, in order to offer quantified discussion on power consumption, this appendix explains the characteristics of the parametron using figures in Goto's IRE paper [9] for curious readers who are experts in electrical engineering.
The fundamental details are all described in Goto's paper [9]. This document adds a few more explanations to that paper to make the discussion on power easier to understand. For details, please refer to [9].
We will briefly describe the electrical properties of the parametron in more detail than in the main text and then discuss power consumption issue.
Parametric Oscillation
The very simplistic schematic diagram of parametron is given in Figure-1 in main text, which is again quoted here for convenience.
Appendix-VI Figure 1 Schematic Diagram of Parametron
(same as Figure 1 in the main text box. Source: Information Processing Society of Japan,
Available online at https://museum.ipsj.or.jp/en/computer/dawn/0007.html
The Parametron generates an oscillation at half the frequency of an external excitation oscillation (f). This is based on the principle of parametric oscillation.
Appendix-VI Figure 2 Oscillation of Parametrons
Taken from Figure 4 of Goto's [9].
Seed Signal and Amplification Effect
In the case of a parametron, based on theoretical analysis and actual observations, it is known that the phase of parametric oscillation can either be out of phase or in phase with the excitation oscillation. It is known that there are cases where the phase is not out of phase (synchronous) and cases where the phase is exactly half a period out of phase. This phase shift is mapped to logical 0 and 1, enabling the use of parametrons as logic circuit elements.
Whether the phase shift is 0 or π was determined by the initial seed signal.
The parametron amplifies the seed signal through nonlinear amplification, ultimately producing a stable signal. It is an important property of the parametron that it amplifies the seed signal while maintaining the phase shift. If a seed signal is not provided, the phase shift is determined randomly based on noise input. For the parametron to function as a logic element, multiple signals from the preceding parametrons are added as seed signal.
We have to realize that, for the final phase of the oscillation determined by amplifying the input signal through nonlinear operation to be clearly defined, several full cycles of the oscillation are necessary for the oscillation to stabilize into a stable shape. The number of cycles required for the oscillation to stabilize becomes the operational speed limit of the parametron.
Note: To generate several full cycles of oscillation, the excitation oscillation must continue during that time. Since the excitation oscillation has twice the frequency of the parametron oscillation, twice as many number of excitation current oscillations are required as the number of parametron oscillations until stabilization.
In the case of parametron elements as logic circuits, multiple outputs
from the preceding parametrons are analogically added to form
the seed signal to the next parametron to be amplified.
Majority principle
The sum of signals from multiple parametrons is obtained by winding the signal lines in the same direction around a single core. This allows the result of the analog signal sum to be produced. In the device developed by Goto, an additional toroidal core is used, and the input signals are wound in the same direction around it to generate magnetic flux due to the analog sum of the signals, which is then transmitted to the ferrite cores where parametric oscillation occurs as initial seed signal.
Appendix-VI Figure 3 Diagram of a typical parametron
Note the use of a coupling toroidal core (a transformer) that accepts input signals (wound only one turn)
The exciting signal is applied in different directions against F1 and F2 ferrite cores.
There is a register for damping and coupling.
The rightmost core is for the input of the next parametron stage.
The same physical behavior can be realized using the “binocular type core”
developed later to decrease the power consumption by reducing the
volume of ferrite core that is magnetized by the exciting current AC signal.
Appendix–VI Figure 4 diagram of a Binocular type core
Figure-20 of Goto's paper [9]
Note: The different direction of exciting signal to the first and
second parts of the core is automatically realized by the wiring in
the figure.
The sum of the analog signals serving as seed signals performs a majority decision. When the phase of the signals is offset by 0 or π, the voltages become opposite (logically interpreted as 0 or 1, but as + versus – in voltage terms). If the majority of inputs to the parametron are “+,” the sum of the analog signals is “+”; if the majority are “–,” the sum is “–.” The result of this analog signal sum (phase difference) is amplified by the parametric oscillator to produce a stable oscillation, which reflects the majority decision result of the seed signal's +/- (phase difference). This signal, reflecting the majority decision operation, is transmitted to the next stage of the parametric oscillator and amplified. This is the principle of parametric logic.
An important feature of parametric elements using ferrite cores is that they have an amplification effect on the input and generate a frequency that is half that of the excitation oscillation, with two states separated by exactly π in phase.
Direction of propagation of electrical signal vs. logical information
There is another point that was omitted from the main text to avoid complicating the description.
Careful analysis of the operation of the parametron reveals that there is no specific direction of signal propagation. In other words, although the description so far has implied a distinction between input and output, in reality, the oscillating signal is transmitted indiscriminately from one ferrite core to another connected to it. The term “seed signal” refers to the signal from a parametron that is already oscillating, which is applied to a parametron that has not yet begun to oscillate at that point and is about to initiate its operation.
When two connected parametric elements are oscillating simultaneously, there is no distinction between input and output, meaning that there is no longer a seed signal, making it difficult to operate as a logic circuit.
Utilization of three excitation current signals operating at different timings
The method adopted by Goto to introduce the direction of logic signal propagation involves applying three excitation signals oscillating at different phases to parametrons in logical stages 1, 2, and 3 (and similarly, 3n+1, 3n+2, and 3n+3 logical operation steps), enabling the signals to be processed through majority logic while progressing to the next logical step. This prevents signals from returning to the previous stage parametrons and causing malfunctions.
Appendix–VI Figure 5 Exciting current of three groups, I, II and II
Figure-20 of Goto's paper [9]
Explanation: First, consider three excitation signals that operate at three different timings as shown in Appendix-VI Figure-5. When the first excitation signal is operating, the other two are not operating. When the first excitation signal stops, the second signal starts operating (actually slightly earlier). Then, when the second excitation signal stops, the third signal starts operating, again slightly earlier. When the third excitation signal stops, the first signal starts operating again.
The parametron circuit is designed so that the parametron operates with three excitation signals in accordance with the logical sequence of steps. The logical step 1 operates with the excitation signal I, the step 2 with the signal II, and the step 3 with the signal III, etc.
Figure Appendix VI Figure 6 A parametron delay circuit.
Description: Please refer to Figure 6. The first-stage parametron
operates with the phase I excitation signal. At this point, the
next-stage parametron has not received the excitation signal and is
not oscillating. The phase II excitation signal begins to
strengthen. The parametric oscillation of the second-stage parametron
begins. The signal from the first-stage parametron is transmitted to
the second-stage parametron, becoming one component of the analog sum
of the input signals to the second-stage parametron. (In this case,
this is a simple diagram of delay line as an example. Thus there is
only one input to the next stage parametron.)
This majority decision result is amplified, and the second-stage parametric
oscillation stabilizes after a dozen or so cycles with its phase
finally stabilized. The phase I excitation signal then weakens, and the oscillation
of the first-stage parametron stops.
Furthermore, the excitation signal for phase III begins to strengthen. Similarly, the signal from the second-stage parametric oscillator is transmitted as the majority decision input to the third-stage parametric oscillator. The third-stage parametric oscillation stabilizes after a dozen cycles as a result of this process. The excitation signal for phase II weakens and disappears. The excitation signals I, II, and III are then applied in a cyclic manner, and the entire parametric element circuit performs logical operations.
Figure 6 shows a simple delay circuit, but Figure 7 shows a circuit that sends the majority rule result of (x, y, z) inputs to the next step.
Appendix VI Figure 7 Majority of (x, y, z)
Figure 8 of Goto's [9]
Note that the first stage is driven by exciting current with phase I, and the next stage is driven by exciting current with phase II.
Number of excitation signals operating at different timings: 3
At least three excitation signals operating at different timings are required. If only two are used, even if the majority decision operation is performed when the signal moves from the first stage to the second stage, the signal representing the logical operation result returns to the first stage when the first stage is excited again, preventing correct logical operation. Therefore, at least three different phases are necessary. However, there is no particular advantage to increasing the number to number 4 or beyond.
Parametric oscillation energy dissipation: Presence of resistor
The energy of the oscillation of the excited parametron dissipates and disappears through the connected resistor when the excitation signal is no longer applied. Although the simplified principle diagram in Figure-1 does not show the resistor, it is necessary for this purpose. The presence of resistor can be seen in Figure-3 and Figure-4 of this Appendix VI. It can be also seen in the photo of a very rare PC-1 board. Goto mentioned that no photo of PC-1 and the circuit remained in his hands in 1975 [6]. But, his colleagues preserved the photos. This board in the picture I quote below seems to be that of the board which Dr. Kei Hiraki, then an undergraduate student at the Takahasi Goto laboratory, picked up from the trash bin when Takahasi retired from the University of Tokyo and some curiosities left behind in the laboratory rooms were dumped in annual cleanup. Dr. Hiraki preserved it and used it in the news of the faculty of science, the University of Tokyo.
Photo - Parametron board from PC-1
Source: from the cover of The Faculty of Science Bulletin, January 2009 issue, Vol 40, No. 5, Faculty of Science, the University of Tokyo, January 2009
Some of the oscillation energy is also dissipated as heat through the
ferrite core. Due to hysteresis in the magnetic properties of the
core, heat is generated when alternating current induces magnetism. As
a result, the ferrite core becomes hot. However, in the case of the
Parametron computer PC-1, no special cooling was required. The faster
PC-2 generated more heat and required cooling. The room where the PC-2
was installed was cool, and graduate students studied there in the summer.
Power, excitation signal frequency, and Parametron operating speed
Now, with the above explanation, we can finally discuss the operational frequency and power consumption of the Parametron quantitatively.
The power supply for the Parametron circuit is primarily provided by an external excitation signal.
As can be understood from the operational principles explained above, a power supply device capable of generating excitation signals with three distinct phases is required.
Designers of circuits using parametrons can determine the number of cycles required for parametric oscillation to stabilize based on experience. Increasing the number of cycles stabilizes operation but slows down logical operations. Decreasing the number of cycles speeds up logical operations but increases the risk of analog-like malfunctions. A description stating that 10 to 20 cycles were used can be found in Ishibashi's memoirs. (“TOYOTA and parametron electronic computer FACOM 202,” Yoshihiro Ishibashi, available online at: https://www.toyotariken.jp/_/media/page/about/research-report/pdf/Toyota-Report_No.77_62.pdf) Depending on the number of times, it is necessary to switch the activity length of the excitation signal with three phases. This can be considered as the gate delay of the parametron logic circuit.
The clock (clock) of various Parametron computers listed in the table in Goto's paper [9] (cited in Appendix V) is considered to be the switching frequency of the excitation frequency (where kc denotes kilohertz (kHz)). The exciting frequency is the frequency of the excitation current (where mc denotes megahertz (MHz)). In the case of PC-1, the excitation signal frequency is 2 MHz, and the parametron operating frequency is 15 kHz, meaning that three excitation signals are switched at 15 kHz. This results in three groups of parametrons being excited and stopped at a period of 45 kHz.
If the proposer's calculations are correct, the parametron oscillates approximately 22 ( 2Mhz/15KHz/2/3) times after excitation before stopping (note that the parametric oscillation of the parametron occurs at half the period of the external excitation signal). In the case of PC-2, the frequencies are 6 MHz and 100 kHz, and it appears that the parametron is oscillating approximately 10 times after excitation until it stops. Note that this number includes oscillations with incomplete waveforms during the rise phase.
Barrier against increased operating speed and reduced power consumption
To increase the operating speed of the parametron, the following is necessary.
The switching frequency of the excitation signals for the three phases must be increased. Note that this also serves as the power supply. It is quite difficult to create three output power supplies that operate at high output and high frequencies in three phases. Currently, the control circuit can be made using ICs and power transistors, but this was even more difficult in the past.
Furthermore, to achieve phase stability with a sufficient number of oscillations (10–30) during the shorter operation times of the three phases due to high-speed operation, it is necessary to increase the frequency of the parametric oscillation, which requires increasing the frequency of the excitation oscillation itself. Stable operation of a high-output, high-frequency power supply was difficult in the past (and may still be today). Furthermore, increasing the frequency of parametric oscillations leads to increased heat generation in ferrite cores with hysteresis and in resistors used for damping. Since ferrite cores change their magnetic properties with temperature, robust cooling is necessary.
While the parametric oscillator was faster than an electromechanical relay, achieving the speed required to compete with vacuum tubes and the emerging transistors posed challenges such as the difficulty of constructing a high-frequency AC power circuit with three phases, which was difficult to implement, and issues related to heat dissipation due to increased power consumption.
These challenges were some of the reasons for the decline of the parametron due to the advancement of transistors.
Supporting materials (supported formats: GIF, JPEG, PNG, PDF, DOC): All supporting materials must be in English, or if not in English, accompanied by an English translation. You must supply the texts or excerpts themselves, not just the references. For documents that are copyright-encumbered, or which you do not have rights to post, email the documents themselves to ieee-history@ieee.org. Please see the Milestone Program Guidelines for more information.
Please email a jpeg or PDF a letter in English, or with English translation, from the site owner(s) giving permission to place IEEE milestone plaque on the property, and a letter (or forwarded email) from the appropriate Section Chair supporting the Milestone application to ieee-history@ieee.org with the subject line "Attention: Milestone Administrator." Note that there are multiple texts of the letter depending on whether an IEEE organizational unit other than the section will be paying for the plaque(s).
Please recommend reviewers by emailing their names and email addresses to ieee-history@ieee.org. Please include the docket number and brief title of your proposal in the subject line of all emails.
