Forgotten generation of relay computers

Our previous article described the heyday of automatic telephone switches, which were controlled using relay circuits. This time we want to talk about how scientists and engineers developed relay circuits in the first - now forgotten - generation of digital computers.

Relay in zenith

If you remember, the relay is based on a simple principle: an electromagnet operates with a metal switch. The idea of ​​a relay in the 1830s was independently proposed by several naturalists and entrepreneurs in the field of telegraph business. Then in the middle of the XIX century, inventors and mechanics turned the relay into a reliable and irreplaceable component of telegraph networks. It was in this area that the life of the relay reached its zenith: it was miniaturized, and generations of engineers created a myriad of constructions, formally learning mathematics and physics.

At the beginning of the 20th century, not only automatic switching systems, but practically all the equipment of telephone networks contained some types of relays. One of the earliest uses in telephony refers to 1870, in hand-held switches. When the subscriber twisted the handle of the phone (magneto handle), a signal was sent to the telephone exchange, including a blender. The blender is a relay, when triggered, a metal valve fell on the switching table at the telephone operator, which indicated an incoming call. Then the young lady-operator inserted the plug into the connector, the relay was reset, after which it was possible to raise the flap again, which the electromagnet held in this position.

By 1924, as Bell's two engineers wrote, a typical handheld telephone exchange served about 10,000 subscribers. Its equipment contained 40–65 thousand relays, whose total magnetic force was “sufficient to lift 10 tons”. In large telephone exchanges with machine switches, these characteristics were multiplied by two. Throughout the US telephone system, many millions of relays were used, and their number increased steadily with the automation of telephone exchanges. One telephone connection could serve from a few to several hundred relays, depending on the number and equipment of the telephone stations involved.

The factories of the company Western Electric, the production structure of Bell, produced a huge range of relays. Engineers have created so many modifications that the most sophisticated dog breeders or pigeon hunters would envy this diversity. The speed of operation and sensitivity of the relay were optimized, the dimensions decreased. In 1921, Western Electric produced almost 5 million relays of one hundred basic types. The most popular was the universal Type E relay, a flat, almost rectangular device, weighing several tens of grams. For the most part it was made of stamped metal parts, i.e. it was technologically advanced in production. The case protected contacts from dust and induced currents from neighboring devices: usually relays were mounted close to each other, in racks with hundreds and thousands of relays. A total of 3,000 Type E variants were developed, each of which differed in winding and contact configurations.

Soon, these relays began to be used in the most complex switches.

Coordinate switch

In 1910, Gotthilf Betulander, an engineer at Royal Telegrafverket, a state-owned corporation that controlled most of the Swedish telephone market (for almost all decades), came up with an idea. He believed that he could greatly improve the efficiency of Telegrafverket’s operations through the construction of fully relay-based automatic switching systems. More precisely, on the relay matrices: lattices of steel rods connected to the telephone lines, with the relay at the intersection of the rods. Such a switch should work faster, more reliably and be easier to maintain compared to systems based on sliding or rotating pins.

Moreover, Betulander invented that it was possible to isolate the parts of the system responsible for selection and connection into independent relay circuits. And the rest of the system should only be used to set up a voice channel, and then be released to service another call. That is, Betulander came to the idea, which was later called the "common control" (common control).

He called the scheme storing the incoming call number “recorder” (another term is register). A scheme that finds in the grid and “marks” an available connection is called a “marker”. The author has patented his system. Several such stations appeared in Stockholm and London. And in 1918, Bethulander found out about American innovation: a coordinate switch created by Bell engineer John Reynolds five years ago. This switch was very much like the development of Betulander, but it used n + m relays to serve n + m matrix nodes, which was much more convenient for the further expansion of telephone exchanges. When the connection was established, the holding rail clamped the “fingers” of the piano strings, and the selector rail moved along the matrix to connect with another call. The following year, Betulander introduced this idea into the design of his switch.

But most engineers considered the creation of Betulander strange and unnecessarily complicated. When it came time to choose a switching system to automate the networks of the largest Swedish cities, Telegrafverket preferred the design developed by Ericsson. Betulander switches were used only in small telephone exchanges in rural areas: the relays were more reliable than the motorized automation of Ericsson switches and did not require service technicians at each station.

However, American telephone engineers had a different opinion on this matter. In 1930, Bell Labs specialists arrived in Sweden and were "very impressed with the parameters of the coordinate switching module." When they returned, the Americans immediately began working on what later became known as the “coordinate system No. 1”, replacing panel switches in large cities. By 1938, two such systems were installed in New York. Soon they turned into standard equipment for city telephone exchanges, until more than 30 years later they were replaced by electronic switches.

The most interesting component of the coordinate switch number 1 was the new, more complex marker developed in Bell. It was intended to search for a free route from the caller to the callee through several coordinate modules connected with each other, thereby creating a telephone connection. The marker also had to test each connection for a “free” / “busy” state. This required the use of conditional logic. As historian Robert Chapuis wrote:

The choice is conditional, because the free connection is retained only if it provides access to the coordinate rail, which, as an exit, has a free connection with the next level. If several sets of connections satisfy the desired conditions, then the “preferential logic” chooses one of the [existing] smallest connections ...

Coordinate switch is a great example of mutual enrichment of technological ideas. Betulander created his fully relay switch, then improved it with a Reynolds switching matrix and proved the performance of the resulting construction. AT & T engineers later reworked this hybrid switch, improved it, and created coordinate system No. 1. This system then became a component of two early computers, one of which is now known as a milestone in the history of computing.

Mathematical calculations (Mathematical labor)

To understand how and why relays and their electronic cousins ​​helped make a revolution in computing, we need a brief excursion into the world of mathematical computing. After it, it becomes clear why there was a latent demand for optimization of computational processes.

By the beginning of the 20th century, the whole system of modern science and engineering was based on the work of thousands of people who carried out mathematical calculations. They were called computers (computers) [ To avoid confusion, hereinafter the term calculators will be used. - Note. per. ]. Back in the 1820s, Charles Babbage created a difference machine (although his apparatus had ideological predecessors). Its main task was to automate the construction of mathematical tables, for example, for navigation (calculation of trigonometric functions by polynomial approximations at 0 degrees, 0.01 degrees, 0.02 degrees, etc.). Also, astronomy was in great demand for mathematical calculations: it was necessary to process raw results of telescope observations in fixed regions of the celestial sphere (and depending on the time and date of observations) or to determine the orbits of new objects (for example, Halley's comet).

Since the days of Babbage, the need for computers has grown many times. Electricity companies needed to understand the behavior of transmission systems with extremely complex dynamic properties. Guns made of Bessemer steel, capable of throwing projectiles beyond the horizon (and therefore, due to direct observation of the target, they were no longer induced), demanded more and more accurate ballistic tables. New statistical tools, which implied a large amount of mathematical calculations (for example, the least squares method), were increasingly used both in science and in the growing state apparatus. In universities, government offices, and industrial corporations, computing departments appeared that usually recruited women.

Mechanical calculators only simplified the task of computing, and did not solve it. Calculators accelerated arithmetic operations, but any complex scientific or engineering tasks required hundreds or thousands of operations, each of which the computer (the person) had to perform manually, carefully recording all intermediate results.

The emergence of new approaches to the problem of mathematical calculations has contributed to several factors. Young scientists and engineers, who had painfully cheated their tasks at night, wanted to rest their hands and eyes. Project managers were forced to lay out more and more money for wages to numerous calculators, especially after the First World War. Finally, many advanced scientific and engineering tasks with great difficulty succumbed to manual computation. All of these factors led to the creation of a series of computers, the work on which was carried out under the leadership of Vanivar Bush (Vannevar Bush) - an electrical engineer from the Massachusetts Institute of Technology (MIT).

Differential analyzer

Up to this point, history has often been impersonal, but now we will talk more about specific people. Glory has bypassed the creators of the panel switch, the Type E relay and the coordinate marker circuit. About them not even biographical jokes. The only publicly available evidence of their life is the fossil remains of the machines they created.

Now we can get a deeper understanding of people and their past. But we will no longer meet those who worked hard in the attics and workshops in their homes - Morse and Weil, Bell and Watson. By the end of the First World Era, heroic inventors were almost over. Thomas Edison can be considered a transition figure: at the beginning of his career he was a mercenary inventor, and by its end he became the owner of the “factory of inventions”. By that time, the development of the most notable new technologies had become the domain of organizations - universities, research divisions of corporations, government laboratories. The people we will talk about in this section, just belonged to similar organizations.

For example, Vanivar Bush. He arrived at MIT in 1919, when he was 29 years old. After a little more than 20 years, he became one of the people who influenced US participation in World War II, and helped increase government funding, which forever changed the relationship between government, academia and the development of science and technology. But within the framework of this article, we are interested in a series of machines that have been developed in the Bush laboratory since the mid-1920s and were intended to solve the problem of mathematical calculations.

MIT, recently relocated from central Boston to the Charles River Embankment in Cambridge, was closely associated with industry needs. Bush himself, in addition to his professorship, had financial interests in several enterprises in the field of electronics. So you should not be surprised by the fact that the problem that led Bush and his students to work on a new computing device arose in the energy industry: it took to simulate the behavior of the main transmission lines under peak loads. Obviously, this was only one of many possible applications of computers: tedious mathematical calculations were carried out everywhere.

Bush and his colleagues first built two machines, which were called product integraphs. But the most famous and successful MIT machine was another one - a differential analyzer , completed in 1931. He solved problems with the transmission of electricity, calculated the orbits of electrons, the trajectory of cosmic radiation in the Earth’s magnetic field and much more. Researchers around the world who needed computing power created dozens of copies and variants of a differential analyzer in the 1930s. Some - even from Meccano (English equivalent of American children's designers of the brand Erector Set ).

The differential analyzer is an analog computer. Mathematical functions were calculated using rotating metal rods, the rotation speed of each of which reflected some quantitative value. The motor actuated an independent rod — a variable (usually it was time), which, in turn, rotated other rods (different differential variables) through mechanical connections, and a function was calculated based on the input rotation speed. The results of the calculations were drawn on paper as curves. The most important components were the integrators - the wheels, rotating discs. Integrators could calculate the integral of the curve without tedious manual calculations.

Differential analyzer. The integrated module - with the lid lifted, on the side of the window there are tables with the results of calculations, and in the middle - a complex of computing rods

None of the analyzer components contained discrete switching relays or any digital switches. So why do we talk about this device? The answer is given by the fourth car of the family.

In the early 1930s, Bush began courting the Rockefeller Foundation to receive funding for the further development of the analyzer. Warren Weaver, head of the natural sciences department of the foundation, could not be convinced at first. Engineering was not his responsibility. However, Bush advertised the limitless potential of his new machine for scientific applications - especially in mathematical biology, Weaver's favorite project. Bush also promised numerous improvements to the analyzer, including "the ability to quickly switch the analyzer from one problem to another, like a telephone switchboard." In 1936, his efforts were rewarded with a grant of 85 thousand dollars allocated for the creation of a new device, which was later called the Rockefeller differential analyzer.

As a practical calculator, this analyzer was not an outstanding breakthrough. Bush, who became MIT's vice president and dean of the engineering department, could not devote much time to managing the development. In fact, he soon withdrew, taking up duties as chairman of the Carnegie Institution in Washington. Bush sensed the approach of war, and he had several scientific and production ideas that could serve the needs of the armed forces. That is, he wanted to be closer to the center of forces, where he could more effectively influence the solution of certain issues.

At the same time, the technical problems dictated by the new design were solved by the laboratory staff, and soon they began to be distracted to work on military tasks. The Rockefeller machine was completed only in 1942. The military found it useful for mass production of ballistic tables for artillery. But soon this device was eclipsed by purely digital computers - the numbers were not as physical quantities, but abstractly, using switch positions. It just so happened that the Rockefeller analyzer itself used quite a few of these switches, consisting of relay circuits.

Shannon

In 1936, Claude Shannon was only 20 years old, but he already graduated from the University of Michigan with a bachelor's degree in two specialties: electrical engineering and mathematics. At MIT, he was led by a leaflet pinned to a bulletin board. Vanivar Bush was looking for a new assistant to work on a differential analyzer. Shannon without hesitation filed an application and soon began working on fresh problems, and only after that the new device began to take shape.

Shannon was not at all like Bush. He was neither a businessman, nor a builder of an academic empire, nor an administrator. All his life he loved games, puzzles and entertainment: chess, juggling, mazes, cryptograms. Like many men of his era, during the war, Shannon devoted himself to a serious matter: he held a position in Bell Labs under a government contract that protected his frail body from military conscription. His research on shooting control and cryptography at that time led, in turn, to the emergence of a fundamental work on information theory (we will not touch it). In the 1950s, when the war and its aftermath subsided, Shannon returned to teaching at MIT, spending his free time on entertainment: a calculator that works exclusively with Roman numerals; the machine, with the inclusion of which a mechanical hand appeared from it and turned off the machine.

The structure of the Rockefeller machine, which Shannon encountered, logically remained the same as that of the analyzer of 1931, but it was built from completely different physical components. Bush realized that rods and mechanical gears in old machines reduced the efficiency of their use: to perform the calculations, it was necessary to set up the machine, which took many man-hours of work of qualified mechanics.

The new analyzer has lost this flaw. At the core of its design was not a table with rods, but a coordinate switch — an extra prototype donated by Bell Labs. Instead of transferring power from the central shaft, each integral module was independently driven by an electric motor. To configure the machine to solve the new problem, it was enough just to configure the relay in the coordinate matrix to connect the integrators in the desired sequence. The punched tape reader (borrowed from another telecommunications device, a roll teletype) read the machine configuration, and the relay circuit converted the signal from the tape into control signals for the matrix — it was like setting up a series of telephone calls between integrators.

The new machine was not only much faster and easier to set up, but also worked faster and more accurately than its predecessor. She could solve more complex problems. Today, this computer can be considered primitive, even extravagant, but then it seemed to observers to be some great — or, perhaps, terrible — working mind:

In fact, this is a mathematical robot. An automaton driven by electricity, created not just to remove the burden of heavy computation and analysis from the human brain, but also to lash out and solve mathematical problems beyond the power of mental decision.

Shannon concentrated on converting data from a paper tape into instructions for the “brain,” and the relay circuit was responsible for this operation. He drew attention to the correspondence between the structure of the scheme and the mathematical structures of Boolean algebra, which he studied in the senior class in Michigan. This is an algebra whose operands were TRUE and FALSE , and the operators are AND, OR, NOT , etc. The algebra corresponding to logical assertions.

After spending the summer of 1937 working at Bell Labs in Manhattan (an ideal place to think about relay circuits), Shannon wrote his master's thesis entitled "Symbolic Analysis of Relay and Switching Circuits" (A Symbolic Analysis of Relay and Switching Circuits). Along with the work of Alan Turing, created a year earlier, Shannon's dissertation formed the foundation of the science of computers.

In the 1940–1950s, Shannon built several computational / logical machines: a calculator with the Roman calculus THROBAC, a machine for chess endgames, and also Theseus, a maze through which the electromechanical mouse went (in the photo)

Shannon discovered that the system of equations of propositional logic can be directly mechanistically transformed into a physical circuit of relay switches. He concluded: "In fact, any operation that can be described in a finite number of steps using the words IF, AND, OR , etc., can be automatically performed using a relay." For example, two controlled relays of a switch connected in series form a logical AND : current will flow along the main wire only when both electromagnets are activated to close the switches. At the same time, two relays connected in parallel form an OR : current flows in the main circuit activated by one of the electromagnets. The output of such a logic circuit can, in turn, control the electromagnets of other relays in order to obtain more complex logical operations like (A & B) or (B & D).

Shannon completed his thesis with an application with several examples of schemes created by his method. Since the operations of Boolean algebra are very similar to arithmetic operations in the binary system (that is, using binary numbers), he showed how to assemble an “electrical adder in a binary system” from the relay — we call this the binary adder. A few months later, one of the scientists at Bell Labs made such an adder on the kitchen table.

Shtibitz

George Stibitz (George Stibitz), a researcher from the mathematical department of the headquarters of Bell Labs in Manhattan, a dark November evening in 1937 brought home a strange set of equipment. Dry battery cells, two small lights for hardware boards and a pair of flat Type U relays found in the trash can. Having added several wires and some trash, he assembled a device that could add two one-digit binary numbers (represented by the presence or absence of input voltage) and output a two-digit number using light bulbs: unit - on, zero - off.

Stibitsa binary adder

Stibitsa, a physicist by training, was asked to evaluate the physical properties of relay magnets. Previously, he had no experience with relays at all and therefore began by studying their use in Bell phone circuits. Soon, George noticed similarities between some schemes and arithmetic operations with binary numbers. Intrigued, he gathered his side project at the kitchen table.

At first, Shtibitsa’s pampering with the relay aroused little interest among Bell Labs management. But in 1938, the head of the research group asked George if his calculators could be used for arithmetic operations with complex numbers (for example, a + b i , where i is the square root of a negative number). It turned out that several computational divisions at Bell Labs were already groaning because they constantly had to multiply and divide such numbers. Multiplying one complex number required four arithmetic operations on a desktop calculator, division - 16 operations. Shtibitz said that he could solve the problem and developed a machine diagram for such calculations.

The final design, which telephone engineer Samuel Williams embodied in metal, was called Complex Number Computer - or Complex Computer for short - and was launched in 1940. For the calculations, 450 relays were used, the intermediate results were stored in ten coordinate switches. Data was entered and received using a roll teletype. The departments of Bell Labs have installed three such teletype, which indicates a large need for computing power. Relay, matrix, teletype - in all respects it was a product of the Bell system.

The finest hour Computer Complex struck September 11, 1940th. Shtibitz presented a computer report at a meeting of the American Mathematical Society at Dartmouth College. He agreed that a teletype with telegraph connection to the Complex Computer in Manhattan, 400 kilometers away, would be installed there. Those interested could go to the teletype, enter the conditions of the problem on the keyboard, and see how the teletype magically prints the result in less than a minute. Among those who experienced the novelty were John Mauchly and John von Neumann, each of whom would play an important role in the continuation of our story.

The meeting participants saw a brief glimpse of the future world. Later, computers became so expensive that administrators could no longer allow them to stand idle while the user was scratching his chin in front of the control console, wondering what to type next. For the next 20 years, scientists will begin to think about how to create general-purpose computers that will always expect you to enter data into them, even while working on something else. And then it will be another 20 years until this interactive mode of calculations becomes in the order of things.

Stiitz for the Dartmouth Interactive Terminal in the 1960s. Dartmouth College was a pioneer in interactive computing. Shtibitz became a college professor in 1964

It is surprising that, in spite of the tasks that it solves, Complex Computer by modern standards is not a computer at all. He could perform arithmetic operations with complex numbers and, probably, solve other similar problems, but not general-purpose problems. . . , .

Bell Model II, Model III Model IV (Complex Computer, , Model I). , , . .

, Ballistic Calculator ( — Model III) . 1944- , . 1400 , . , — . , , . Bell (hunting circuits), / . , Model III, , 25—40 .

Bell Model III

Model V . . , Model III. 9 , . . . ( , ) , . .

: 1937-

1937 . . Bell Labs . — , — , , .

Proceedings of the London Mathematical Society « » (On Computable Numbers, With an Application to the Entscheidungsproblem). : , , -. , , . , , . — , . , .

1937- (Howard Aiken) . - . , . , , .

Nippon Electric Company (Akira Nakashima) 1935 . , 1938- , .

(Konrad Zuse), , , , . — V1, , , (Helmut Schreyer).

, , — .

1937- . , , . . — .

1944- Harvard Mark I / IBM Automatic Sequence Controlled Calculator (ASCC), . - : . Mark I / ASCC , IBM. . Mark II 1948- , — 13 .

, . V4, , Bell Model V, . - , . 1950- , 20 . Fujitsu, , .

. — (ENIAC). , , . , , , . , — , .

. . , . . , . , . , , .