Forrester and his Whirlwind

Background to the creation of tube computers - Whirlwind ("Whirlwind")

In 1943-1944, a need arose for the development of a universal simulator that could simulate the flight of an aircraft in real time. This would give pilots the opportunity to improve their skills, and designers to study the effect of changes in the possible tactical and technical characteristics of new models. Until that time, there were such simulators (based on analogue electromechanical technology), but they were not universal, that is, they were suitable only for a certain aircraft design.



Creating a universal simulator would help save on design and training.

The Massachusetts Institute of Technology has become one of the centers where American air defense systems have been improved. At the end of 1944, work began on the creation of such a simulator at MIT; the laboratory of servomechanisms became the project executor. Jay Forrester led the ASCA project, and Robert Ryvers Everett, MSc in Electrical Engineering, became an assistant.


airplane simulator

Together with a group of employees, they worked for six months to create an electromechanical analog simulator. Real-time electronic analog technology gave low accuracy of calculations, was not able to simultaneously solve hundreds or more differential equations with many variables, and, therefore, did not correspond to the customer requirements, and simple operations required long fuss with punched cards or punched tapes.




It was necessary to look for a new solution to the problem.

Perry O. Crawford, Forrester’s colleague, “contributed much to the birth of the digital VT machine.” ... he told me about the Mark I Harvard machine and ENIAC, which was still in the design stage. Perry was a very open person, devoid of bureaucratic manners and easily communicated with naval commanders, although he himself was a civilian, he inspired them with the idea that digital computers in the future should be used at command posts as the basis of information and management systems ... We owe him a lot to the birth of our machine s. "

Work began on the creation of the machine, which was given the name "Whirlwind" to emphasize its speed. Employees began to develop a draft version of the digital machine (the customer was not notified of such a turn of events). Forrester and Everett familiarized themselves with the structure of Mark I and ENIAC, consulted with their developers, studied the “First Version of the EDVAC Report” (Neumann) and began to sculpt the machine from scratch.

In 1945, Forrester turned to SDD executives. He suggested that they create a simulator not on rough analog computing devices, but on the basis of a digital computer. Complex work was to be done to create a computer that could work in real time.

It was necessary to increase its speed to 20-50 thousand operations per second, use reliable internal memory of high capacity, create software for processing the incoming data stream, and increase the reliability of the machine. The main difficulty was that people who did not have experience in working with digital computers should have dealt with this issue.

Work began on creating an internal memory that would satisfy the requirements of the project, and for this it was necessary to increase the reliability of tube discrete circuits. The main memory of the EDVAC computer, which consisted of 32 mercury ultrasonic delay lines (RULZ), was not suitable for the new project due to slowness. Forrester tried to use gas-filled lamps (“neonki”) and xenon flash lamps (xenon flash lamp) to store one category. In early 1947, English scientists began to use a standard cathode ray tube (CRT) to create internal memory, it made it possible to store one binary digit on its screen and read from it. For four years, MIT employees worked on the design of the tube, the diameter and uniformity of the spot on the tube screen were stabilized, the focus was improved,


RULZ

Work was underway to increase the reliability of lamp circuits. Until that time (as you know) thousands of lamps were used in lamp machines, but the “life” of such a lamp was 500 hours, and this would lead to a malfunction of the machine every couple of minutes. The lamp life has been increased. In 1947, a multiplier was developed, which contained 4 hundreds of lamps, for multiplying two 5-bit binary numbers. Everett wrote "... the device worked continuously for many days, and we checked each result with the correct answer. Of course, it failed, but we noticed that in most cases they occurred at 3 o'clock in the afternoon. It turned out that at that time the cleaner in the next entrance of the building, it included a freight elevator and there was an additional load on the local power grid, which led to malfunctions.



Later, another problem arose that required a solution. The sponsor has changed (OSRD and SDD ceased to exist, and they financed this project), the cost of the project for developing such a computer was constantly increasing. It was planned to create a Whirlwind in two years, with a budget of $ 875 thousand, but the project cost increased to $ 3 million (which amounted to 65% of the total budget of the mathematical department). To justify this expense, Forrester prepared a report for the Pentagon, which outlined where such a host computer could be used. But nevertheless, instead of the requested 1.5 million dollars per year for the project, Forrester and his project were allocated only 250 thousand dollars. The project was saved by a new sponsor - the US Air Force (it happened in 1949 when the Americans learned that an atomic bomb was tested in the USSR). The project to create a computer "Whirlwind" was approved by the Pentagon and allocated the necessary amount for its completion. In March 1951, the machine was completely assembled, debugged and completed the first large program written in assembler by John Gilmore, in April of that year it was put into operation, and Forrester headed the created laboratory of digital computer technology at MIT.

Computer "Whirlwind"

It took almost $ 5 million to develop, based on classic Princeton architecture, but a common bus was used to exchange information between the machine blocks. The diode matrix was used to control the sequence of operations in the computer, the signals from the clock generator were fed to certain inputs of the matrix, and at the same time its output signals opened the keys, with the help of which the code of the desired command was selected, which entered the corresponding register of the control device. The length of the machine word was limited to 16 bits, this included the sign of a fixed-point number, subroutines for working with floating-point numbers and a double-length word.

The machine housed 5,000 lamps (mainly pentodes), 11 thousand crystalline diodes, it consumed 150 kW of energy, the weight of such a computer was 10 tons, and its area was almost 950 square meters. The

whirlwind was a unicast, synchronous operation with a clock frequency of 1 MHz, a machine with internal memory, which contained 32 modified Williams tubes with a capacity of 1K words. A group of 32-bit registers built on mechanical keys and 5 electronic registers served to test the memory.

The punched tape and reader were used to input data and programs into computers. A CRT 40 cm in diameter served as a display. Data was entered into the machine on a tape perforated by a flexor. And the results were displayed on the screen of a CRT or displayed on the same flexor.
The cathode-ray tube was like a television tube, a large electron-vacuum tube in which a beam of electrons, falling on the phosphorus-coated inner surface of the screen, caused it to glow. The speed of the machine was 20 thousand operations per second, the addition operation was performed in 49 μs, and the multiplication in 61 μs.

The Vikhr computer was modernized and the number of lamps increased to 12.5 thousand, and diodes - up to 23.8 thousand, the machine was occupied by two floors in one of the MIT buildings. On one were tape drives and communication devices with objects. Part of the machine, internal memory and control panel are located on another, in the basement there was a power unit (power 150 kW), an air conditioner was mounted on the roof of the building.

Maintaining a computer memory cost $ 32,000 per month (one tube cost about $ 1,000), so Forrester was looking for a CRT replacement. In 1951, the first memory sample was created on ferrite cores, which two years later replaced the memory with a CRT. Computer performance doubled, the addition operation took 8 μs., Multiplication -25.5 μs., Division - 57 μs.


memory on ferrite cores

Five magnetic tape drives (with a capacity of 125 thousand words each and a read speed of 390 words per second) and two magnetic drum drives (with a capacity of 2048 words and a rotation speed of 60 rps, now served as external storage devices) readings - 31 thousand words per second).

Charactron tubes (like displays) were used to play letters, numbers, topographic signs, and other symbols on a fluorescent screen. An opaque plate with a set of microholes in the form of depicted symbols served as a stencil with which the symbols were formed on the screen. The machine software allowed the computer to solve several problems simultaneously, working in time-sharing mode.

Direct memory access (DMA) method : the input data was updated every 15 seconds and recorded on a separate track of one of the magnetic drum drives, after which the unit transferred to the internal memory (the calculation process was continuous).

The whirlwind became the prototype of a number of computers that made it possible to create a powerful US air defense system “SAGE”, a semi-automatic system that can simultaneously process data coming from 23 regional centers in the USA and Canada, while serving a gigantic network of radars and other detectors.



It looked like this: the operator in each district center entered data on the keyboard, looking at the round screens where weather conditions were displayed, the trajectory of the aircraft, the information necessary for the operation of the air defense system. SAGE I / O devices supported continuous telephone line communications between neighboring centers over telephone lines. About 8-12 billion dollars were spent on creating SAGE.


workplace of the operator of the first US national air defense system - SAGE

Work on the Whirlwind project gave invaluable experience to its creators and developers. Many of them took leading positions in many well-known companies: Kenneth Olsen founded the Digital Equipment (DEC) corporation in 1957 - he was engaged in the production of mini-computers.

Forrester himself remained the head of the department until 1956, later devoted himself to researching the dynamics of production, as well as world social and economic processes, and became the founder of a new discipline - system dynamics. Everett and Forrester were awarded the National Medal of Excellence in Engineering, the highest government award for engineers in the United States.