April 1971 Popular Electronics
Table of Contents
Wax nostalgic about and learn from the history of early electronics. See articles
published October 1954 - April 1985. All copyrights are hereby acknowledged.
Albert Einstein declared and proved that time is relative and depends on the observer's perspective. To someone sixty years old, the year 1971 seems like it was just yesterday, but to people born a couple decades ago, it seems like ancient history. Even so, I am taken by surprise when I read a story from a 1971 issue of Popular Electronics that has produced a list of "early computers" and it includes models like the ENIAC and Harvard Mark I. Instinctively, the IBM XT, Apple II, and Packard Bell, and Compaq lines of personal computers (PCs) come to mind. In 1971, there were no PCs. However, if you compile a list of antique computers, then the aforementioned names apply. This article does provide a nice recounting of the evolution of digital computers from Charles Babbage's mechanical Difference Engine through those vacuum tube-based electronic computers.
Battle of the Giant Brains or Electronics Conquers All
By Frank Y. Dill
A Startling Revelation of the Early Days of Digital Computers
Consider for a moment the hardware that goes to make up a digital computer.
Could any way of implementing digital functions be better or more natural than the use of electronic components? Our first reaction is to say "no" - conditioned as we are by a quarter of a century of electronic computers and an $8 billion (annually) computer manufacturing industry. It happens, however, that this has not always been the case. In the late 1930's, the question of how to design a digital computer was very much up in the air.
Frank Y. Dill is a freelance writer specializing in computers and electronics. Above are copies of old engravings of Charles Babbage and his plans tor the Difference Engine.
For more than a century after its theoretical invention, the digital computer was an idea in search of realization through suitable technology. Electronics, while it was used in some limited precision analog devices, was generally overlooked for digital applications. Designers were still thinking of representing discrete quantities in terms of rotating wheels, punched cards, and relays. These devices had shown great promise as far as reliability was concerned and had proven workable in small-scale digital machines. Thus, when the first large-scale digital computers were being designed over 40 years ago, they used non-electronic devices.
The IBM Automatic Sequence Controlled Calculator, also called the Harvard Mark I, was the first electromechanical, general-purpose digital computer. Neat in appearance, it used counter wheels and relays, had relatively small size, required little maintenance, but computations were slow. (Photo courtesy IBM)
Unchallenged success of these non-electronic "giant brains" would have set important design precedents. Fortunately, however, before this could happen construction of an electronic digital computer was begun. Actually, a three-way contest developed between two electromechanical designs and one electronic. The electronic project was at a disadvantage due to the head start enjoyed by its rivals. However, the fate of electronics in computers rested on the success of this project. If this machine were to fail to operate at all or to be not competitive in reliability and overall cost, the future of electronics would be greatly damaged.
Where It All Began. Before witnessing this three-way design race, let's see where the idea of digital computation began. In September 1834, an Englishman named Charles Babbage, hoping to free mathematicians from the drudgery of calculating tables and to rid the tables of the mathematicians' errors, made the first drawings for what he called an "Analytical Engine." These plans contain many of the basic modern computing-principles: punched card input, multiple registers to store intermediate results, automatic sequencing of control, sequencing based on both calculated results and original instructions, and direct mechanical printout of results.
Babbage's device was an extension of the then two-century-old adding and multiplying machines in that it included multiple step calculations whose sequence was controlled by the machine itself. The older machines used a wheel with teeth to advance an adjacent wheel for an arithmetic carry. Similar wheels were to be used in the Analytical Engine to store one thousand numbers of fifty decimal digits each. Calculations and control were to be accomplished by a combination of metal cams, rods, shafts, and levers capable of coupling and decoupling as the program demanded.
Unfortunately, the Analytical Engine was never built. Its construction - like that of twentieth century digital computers - required considerable financial support. The most likely sponsor, the British Government, had already spent £17,000 on an uncompleted, 10-year project of Babbage's called the "Difference Engine." This project had encountered unexpected delays since the machinist's art, lacking modern alloys and mechanical drawing conventions, was unable to fashion Babbage's ideas into metal. So, when Babbage asked the government whether he should continue work on the older project or begin the more powerful and complex Analytical Engine, they, in typical bureaucratic fashion, deliberated for nine years and then said, "neither."
First electronic general-purpose computer was developed by University of Pennsylvania for the Army Ballistic Research Lab. Not mini in size, it solved the speed problem and was predecessor of many electronic computers. (Photo courtesy U. of Penn.)
Unable to obtain financial support, Charles Babbage spent the remainder of his life (until 1871) trying to improve the mechanical technology of the period. The general purpose digital computer remained unbuilt for more than a century after Babbage conceived his plans. However, it was a fruitful period for electromechanical technology and the advances obtained led to the design in the twentieth century of the first large-scale computer.
Babbage is Vindicated. In 1937, Howard H. Aiken, of Harvard University, aware of the school's need for computational facilities, wrote a paper describing the connection between Babbage's ideas and machines then being produced by International Business Machines.1 It turned out that IBM was willing to build such a machine, believing that many of their existing mechanisms could be used with little or no modification.
The project was begun in 1939 at Endicott, N.Y., and was officially called the IBM Automatic Sequence Controlled Calculator. Computer jargon was not tolerant of such long names, of course, so it is now usually referred to as the IBM A.S.C.C. or the Harvard Mark I, depending on the speaker's industrial or academic background.
In this first electromechanical project (as in Babbage's machine), variable numbers were stored on ten-position metal wheels. They were rotated by a shaft connected to a 4-hp motor and were engaged by a magnetically controlled clutch. Unlike the Analytical Engine, numbers were not transferred mechanically, but electrically through a buss. Relays controlled access to the buss and provided for arithmetic carries and borrows, Mechanical wheels and relays were used to implement the mathematical functions.
The Second Project. Proceeding concurrently with the IBM-Harvard effort was a computer design originated by George R. Stibitz of Bell Telephone Laboratories. Not surprisingly, this design relied heavily on existing telephone and teletype devices. The fundamental computing device was the ordinary telephone relay, which was so reliable that it could be expected to operate continually for years without failure. An added advantage was that standard telephone practice already included design and maintenance procedures necessary to keep the computer in order.
A series of six computer models was eventually built by Bell Labs. The success of each provided incentive and design experience which contributed to the next. The Model I, called the Complex Computer, was put into operation in January 1940. It contained 450 relays and was used to perform the arithmetic of complex numbers for the Labs. In the Models II, III, and IV, which followed, 440, 1400, and 1425 relays were used, respectively. Finally in 1944, Bell Labs had gained enough experience to attempt to build a 9000-relay general-purpose computer - the Model V.
Electromechanical technology was used in both of the projects just described. The main difference was that the IBM-Harvard computer used some purely mechanical devices for number storage and for implementing the arithmetic operations. The need for a mechanical drive system however placed severe spatial restrictions on the design. The Bell Labs machines, using only relays, eliminated these space problems, but speed was not greatly improved. This problem was a fundamental one, involving the inertial mass of the moving parts. Even the short time required to close the relay contacts was an important factor.
Solving the Speed Problem. Events leading to a solution of the speed problem were taking place at the same time. Soon after the shut of World War II, the Army's Ballistic Research Laboratory needed more capacity for handling ballistic tables. It had been using a differential analyzer (an analog computer) designed by the Moore School of Electrical Engineering, University of Pennsylvania. Since the Moore School had a larger analyzer, the BRL contracted to use it and the combination developed into what was probably the largest scientific computing group in the world.
Unfortunately, the results were still unsatisfactory. More than a hundred desk calculators were needed to supplement the analog computers. The Army needed, as soon as possible, a better way of producing firing tables, each of which involved 250,000 to 500,000 mathematical operations.
Relay panel from Bell Labs Model I Complex Computer, first of series of 6 relay computers using standard telephone apparatus. (Photo courtesy Bell Labs.)
In the spring of 1943, Dr. John W. Mauchly, a professor at the U. of Pennsylvania, circulated a report which offered a solution to the Army's computing problem. In 1941 he had visited Iowa State College to study the Atanasoff-Berry computer.2 This 300-tube electronic computer was being built to solve algebraic equations. It was never finished but it reinforced Mauchly's belief that electronic high-speed computing devices were feasible. His report advocated building such a machine. An appendix to the report by J, Prosper Eckert, Jr., gave explicit suggestions for implementing Mauchly's ideas in electronic hardware.
The Army's need at that time justified taking a chance on the project and $61,700 was allocated in 1943 for six months of research and development on Project PX, Mauchly's electronic digital computer. The project later became known as ENIAC, an acronym which originally stood for Electronic Numerical Integrator and Computer, though the last word has since been incorrectly reported as Calculator.
During those six developmental months, the most challenging electronic problems ever encountered in a single design were tackled. Since the reliability of a computer is all-important and since the reliability of the whole is no better than that of its individual parts, component reliability was the first major technical consideration.
The most likely component to fail was the vacuum tube. While a single tube had a life expectancy of many thousands of hours, a total of 18,000 tubes were to be used. This meant that the probability of a single tube failure at any particular time was rather high. Prior to this, the largest electronic system was a radar set with 400 tubes, and the problem was important even then.
The solution was to use proven standard tubes and to operate them well below their normal ratings. Filaments were run at 5.7 volts instead of 6.3 and they were rarely turned off to increase their life.3 Plate and screen powers were limited to 25% of rated values.
It is easy to see how the heat generated by all those tubes provided another problem. There were 70,000 resistors and 10,000 capacitors in the system which could be damaged by excessive heat. Of the 150 kW of power consumed by the ENIAC, 80 kW was dissipated by the tubes and another 20 kW was used to drive cooling fans. In addition, each panel of the machine was protected by a thermostat which would shut off that panel if the temperature got over 115°F.
Another problem encountered in the design of the computer was synchronization. Control pulses of two to five microseconds were repeated in cycles. Still other pulses were generated by the computing process. Although the idea of gating is now fundamental to computer design, one of the first applications of a "gating tube" was in the ENIAC computer.4
For storage, the ENIAC used the Eccles-Jordan trigger circuit commonly called a flip-flop. The arrangement of the flip-flops to represent decimal digits was rather unusual in terms of modern design. Ten flip-flops were used to represent a single digit.
This resulted in a very expensive main memory. Eckert estimated that the average cost of storage was $15.00 per decimal digit. By comparison if the largest main memory of a modern computer like the IBM System 370 Model 165 used such a memory, the 3 million bits would cost in excess of one hundred million dollars.
The Race Quickens. When the ENIAC project was just a year old, the complete electromechanical Mark I was unveiled in a public dedication ceremony on August 7, 1944. Its impressive 51-foot, neatly encased exterior was in sharp contrast to the jumble of wire and panels that were to comprise the ENIAC. Actually the Mark I had been tested prior to the start of ENIAC and it had undergone a complete debugging while operating in secret. The age of digital computers had arrived and the Mark I operated around the clock with a down-time record that was impressive by modern standards.
Thus the Mark I posed a real threat to electronic computers. The total need for computers had been greatly underestimated. In fact, a prediction had been made that the entire computing requirements of the U. S. could be satisfied by six computers.5 So the Army might have been justified in cancelling the ENIAC in favor of a Mark I type of machine, especially if unexpected delays or expenses were encountered. Such loss of sponsorship had happened to Babbage a century before.
An even more direct challenge to ENIAC came from the Bell Labs Model V. The same Army group that had contracted for ENIAC, hedging on its gamble, had also ordered one of the two Model V's being built. However, the hedge was not needed since ENIAC was completed and shown to the public in February 1946, after 30 months and $486,804.22 in the making.
Now the time had come to decide which design was best-the Mark I, the Model V, or the ENIAC. The direction computer design was to take lay in the balance. Debate was based on arguments of component reliability, problem set-up time, error-free operation, self-checking ability, and maintenance costs. The builders of ENIAC based their stand on the fundamental advantage of speed, which really meant decreased cost of computing - and this attribute far outweighed ENIAC's disadvantages.
The deciding argument was given by Dr. Mauchly when he said that the life of a computing device should not be based on time alone but on the number of operations it can be expected to perform before failure," A good relay may average 100,000,000 operations before failure. A vacuum tube may be expected to operate reliably at a pulse rate of one operation per microsecond for 5,000 to 10,000 hours. Thus the tube may perform more than 1012 operations compared to 108 for the relay.
Not only did great speed mean cheaper computing, it also offered hope of fulfillment of the cybernetist's dream of real-time control of complicated events. The ENAC could calculate the trajectory of an artillery shell in half the time of the shell's flight.
The strength of ENIAC's success can be seen in changes made by other builders of digital computers. Harvard followed the electromechanical Mark I with a relay Mark II. A twelve-fold increase in speed was achieved. Since this was still no serious challenge to the ENIAC, there followed the Mark III, which got on the electronic bandwagon. The interest of Bell Labs in computer manufacturing was too heavily tied to the use of telephone equipment to convert to electronics. Their contribution to computer technology was to be with the invention of the transistor, not by the direct manufacture of computers.
The Eckert-Mauechly team went commercial and designed the Binac and their small company was absorbed to form the basis of Univac. And what of IBM, the builder of Harvard's Mark I? Suffice it to say that they too recognized the potential of electronic digital computers.
1. "Proposed Automatic Calculating Machine," H. H. Aiken, IEEE Wpectrum, August 1964, p 64.
2. Electronic Digital Systems. R. K. Richards, John Wiley & Sons, New York, 1966.
3. "Electronic Computing Circuits of the ENIAC," A. W. Burks, Proceedings of the IRE, August 1947, p 757.
4. A Report on the ENIAC, Staff of the Moore School, Univ. of Pennsylvania, Philadelphia, 1946.
5. "An Interview with Eckert and Mauchly," H. Bergstein, Datamation, April 1962, p 26.
6. Theory and Techniques for Design of Electronic Digital Computers, Staff of the Moore School, Univ. of Pennsylvania, Philadelphia, 1947.
7. Survey of Automatic Digital Computers, Office of Naval Research, 1953.
Posted February 27, 2019