Go to Antique Computer home page, Go back to Catalogs
|
Please note - the exact source of this document seems uncertain.
It was OCRed from what appears to be a 1981 draft introduction and catalog
for the original "Computer History Museum" in Boston, Mass.
It shows a serious attempt to organize the wide ranging chaotic field of calculating/computing architecture, interacting with the rapidly changing technology. I suspect the organization of computer history, devices, software, ... will never be solved to anyone's complete satisfaction.
On April 22, 2013, Gordon Bell e-mailed - "This came from when DCM started in Marlboro and was called the Digital Computer Museum… which was changed to TCM. "It is a wonderful catalog that, by example, defined what a museum is and does complete with classification and the like. "The Catalog was a publication of all of the artifacts in the museum at the time it was published in 1981. "Gwen or one of her staff in Marlboro produced this—my guess it was a history of science intern, Beth Parkhurst or maybe Paul Ceruzzi. " |
Starting a computer museum today presents analogous problems to those that of the second duchess of Portland. Born in 1714, she was an insatiable shell collector who relied on artistic arrangements until she hired a student of Linnaeus (1707-1778) the father of botanical classification systems. Then the collection was re-arranged according to a taxonomy that would help the viewer understand evolution and relationships. Computing devices -- as beautiful as shells to many people -- need a theory-based classification system.
The purpose of the Digital Computer Museum, to document and preserve the evolution of the computer, from its earliest origin to the latest developments, demands a rigorous, disciplined classification scheme that focuses on the computer itself.
Intuitively, those who have tried to understand computer evolution to consider a tree structure the basis of taxonomies but none have been fully developed for the purpose. (Bell and Newell, 1971; Bell, McNamara and Mudge, 1978; Rogers, 1980; Science Museum, 1975, Sieworek, Bell and Newell, forthcoming). The National Science Foundation tree (Fig. 1) of early computers shows roots and connections but does not name branches. A number of partial systems and some generally agreed upon terms exist for defining a classification system. The Computing Reviews classification system for contents works very well for the extraordinarily broad range of materials including "mathematics, engineering, the natural and social sciences, the humanities, and other fields with critical information about all current publications in any area of the computing sciences." (Sammet, 1980) The work of the AFIPS Taxonomy Committee, Taxonomy of Computer Science and Engineering, provides a confusing semi-lattice covering all possible issues. (AFIPS Taxonomy Committee 1980) Other trees look at only a part of computing. (Weizer 1981, Sammet 1969) The evolutionary model has also resulted in the identification of generations. (Rosen, 1969)
THE GENERATIONS
Within the broadly accepted idea of technological generations, clear criteria can be identified to mark each one. These are:
Generational change is modeled by a series of distinct steps with a new base technology at a significantly different level. The technology base never meets the aspirations and dreams of mankind perceived needs are continually rising. A new base technology only creates a higher takeoff plane. (Maslow, 1943) with each new invention, one or two prominent people often note that it will fulfill all the future computational needs; but each time the aspiration for more computational power only grows. Computers themselves are a technology that may influence a wide spectrum of other phenomena, such as communications and manufacturing. Since the fifties they have become one of the prime movers of technological advance.
A number of ideas and machines are designed and even built out-of-phase with a technology. Ideas that occur before their time often lie dormant in the inventors notebook until the technology evolves to match the idea. Later historians illuminate these early concepts, showing the contemporary entrepreneurs that they are not inventors but only implementers of ancient ideas. In the mid-twentieth century, some letters of Wilhelm Schichard dated 1624 were unearthed. These contained the drawings for the first known digital machine to perform calculations. (Cohen 1980) It is very doubtful that these ideas transmitted from Schichard to his friend Kepler influenced any of the mechanical calculators that were subsequently developed. Similarly, Leonardo's notebooks included drawings for many engineering devices including a calculator, but the mechanical technology at the time had simply not progressed to the necessary degree. The actual inventors that develop a baseline machine for a technology are often tinkerers with that technology and not scholars searching the literature for ideas.
When one or more significant ideas are transformed into a project, then its execution includes inventions that become part of the technological base. A new generation is marked after the project has proven itself, shown not to be a fluke, and adds a new layer to the technological base. The Computer Revolution and beginning of the electronic generation added the technological use of vacuum tubes in orders of magnitude never before experienced in the ENIAC project and the use of magnetic core memory from the Whirlwind project. Since a generation is a convergence of technology and inventions, marking its emergence by a single event is inappropriate. A clustering of events, including patents, publications, and start-up dates that converge are used to justify the selection of a particular year, that then has approximately a five percent accuracy.
The Museum collections begin in 1620 with the beginning of the "Craft Generation". Prior to that information processing was carried out manually, much the same for all of history. Using the product of processing rate and memory size to measure computing power, a 20 order of magnitude increase can be counted since people used stone-based, single register for arithmetic. The most significant gap a revolutionary change occurred with the beginning of the computer era. Before then, memory size was essentially constant at one. Afterwards, computing power began to increase at roughly twice the exponential rate of all past generations.
The name of the generation indicates wide-spread application and use of a predominant technology. The idea that leads to a project triggering a new generation always occurs before the beginning of that generation. The starting date of a generation is marked by the incorporation of a technology into production of a new product, concurrent with significant use. in most cases devices from a previous generation continue to be designed, manufactured and used, often supplying a base on which the new generation is built.
Table I lists the need, use and representative inventions for each of the generations. During the pre-computer generations, evolution was exponential -- each period being half as long as the one preceding it. The rapid change is similar to manufacturing learning curves, whereby a particular unit cost declines by 10-20% each time the cumulative number of units of a given type are built. In the Computer Age, the naming conventions given by industry have been used, and they seem to accurately fit the model.
Generations are primary organization element for the collection and its representation in the catalog. The first four sections present the pre-computer generations. Then the fifth section is devoted to the pioneer computers that spanned the revolutionary bridge. And the remainder of the catalog and collection is open ended; inclusive of all historic generations, i.e., at least one generation removed from the present technological generation or fifteen years old.
THE TAXONOMY
Structuring a taxonomy has paralleled the development of the collection and the exhibits at the Digital Computer Museum. The PMS classification describing the structure of computing structures provides the basic framework. (Sieworek, Bell and Newell forthcoming) PMS allows any computing or software structure to be described hierarchically in terms of eight basic information processing primitives; but does not deal with functional behavior, e.g.., interrupts except those that can be implied by a structure. The PMS system is generally used to provide a structural representation of the components of digital computer systems, in contrast, this taxonomy only encompasses whole computing systems and their antecedents. The following compares the two breakdowns:
| MUSEUM TAXONOMY | CODE | CODE | PMS |
| Memories | M | M | Memories |
| Controls | K | K | Controls |
| Transducers | T | T | Transducers |
| Links & Switches | S | S | Switches |
| . | . | L | Links |
| Calcula | D | D | Data Operation |
| . | . | P | Processor |
| Digital Computer | C | C | Computer |
| Automata | A | . | . |
The criteria defining the tree is the structure of the computing device, neither the organization that made it nor the purpose that it was meant to fulfill. To make an analogy with the animal kingdom, if the bone structure of a horse is that of a fine race horse then it would be classified as such; it would not matter if it were bred by the government and used to pick up garbage, in computing, the EDSAC, built at Cambridge University, is neither classified as an English or university computer but an EDVAC-related machine in the same family as the Maniac and ILLIAC. Thus, differentiation of manufacturers, countries, or by intended users is not part of the taxonomy.
The classical scientific taxonomy system with its seven levels has been adopted to organize and classify all species of relative inventions. The two top levels, kingdom and phylum, are technology and information, respectively. The Museum collection deals with seven classes within the phylum of "information." (Listed above) Each class, like a specie, has life that starts within a given generation, flowers, and then becomes functionally incorporated within another class. Each started, almost as an independent thread, but are now beginning to merge into two dominant classes: computer and automata.
Memory is probably the oldest class starting with early markings on caves and continuing both as significant parts of computers and automata and also as all kinds of human-readable aids to the brain. See ..... for more complete explanations.
Controls reach back to early analog devices, such as the creek water clocks, and have been significant in the mechanization process. At the beginning of the 19th century, card controlled looms gave the notion of sophisticated pattern control to industrial processes via the use of a larger scale memory data-set than hitherto used. Card control ended with a great flourish in the early nineteen sixties with the tabulating machines. Again with the computer on the chip, earlier technologies of control devices are rapidly becoming obsolescent to be replaced by the "on-board" micro-processor.
Transducers take information in one form and put it into another. They are often associated with memory systems, allowing their replication, printing use type (an intermediary form) to duplicate the information into books, the books are then "secondary" memory for people. Transducers really began with the Guttenberg's movable type 'and include teleprinters, tape transports, the telephone, and television sets. These machines are becoming more and more sophisticated and less and less able to be differentiated from computers.
Calculators, other than the manual bead devices, did not develop until the 19th century and have now virtually been displaced by computers. These are the data operators to do the arithmetic in PMS notation. Either calculators are embedded in computers or computers (as they have miniaturized) are embedded in what has traditionally been considered a calculator. The taxonomy of Class Calcula is worked out and explained in the text. (See ....)
Links and switches evolved out of the need for a large number of subscribers all desiring the use of a single system. The first telegraph was a simple device transferring information from a to b. But the growth of the telegraphy and telephony systems in the late nineteenth century created a need to establish elaborate networks linked together with a switching system. The current generation of computers still depend on new methods of linking and switching for cross communication.
Digital Computers emerged in the late forties from a combination of calculator, control, transducer, links and switches, and memory technologies. The section on Pioneer Computers shows the combination of elements that were adopted by the first 16 machines, many of which were patched together with emphases on different Classes. The Class Digital Computer, itself that emerged is certainly more than the sum of these parts, as each has converged and been modified and molded into a new phenomena.
Automata actually started very early with man's desire to replicate himself and their great population explosion took place in the sixteenth century. But only recently, have useful automata been put to work for human purposes and are contemporary to the latest generation of computers. Thus, this class is presently not included per se in the collection; but will be included in the future.
Each of these seven Classes is broken down into Order, Family, Genus, and then identified by Species. Table 2 lists the criteria used for the breakdown of the Classes. Specific descriptions for each of the class are found throughout the catalog.
PRE-COMPUTER GENERATIONS
| . | MANUAL | CRAFT 1620 | MECHANICAL 1810 | ELECTRO-MECHANICAL 1900 |
| NEED | Taxes | Trade Exploration | Industrial Land Division | Census Business |
| USE | Counting | Arithmetic Navigation | Surveying Weaving | Sorting Accounting |
| MACHINE | Abacus | Tables Gunter's Rule | Planimeter Jacquard loom | Hollerith Census Machine Friden calculator |
| . | ELECTRONIC 1950 | TRANSISTOR 1960 |
| NEED | Defense Weather prediction | Space Science |
| USE | Firing Tables Weather Forecasting Management | Simulation Training programmers Accounting |
| MACHINES | Whirlwind, UNIVAC 1, ERA 1101 | CDC 160,IBM 7090, IBM 1401, PDP-1 |
Criteria used in differentiating orders, families, and genus.
| CLASS | ORDER | FAMILY (Technology) | GENUS
| Memory
| Machine interface
| Storage material
| Structure of | access movement Controls
| .
| Complexity
| .
| Transducers
| .
| Phenomena
| .
| Links & | Switches .
| Complexity
| .
| Calcula
| Analog or Digital
| Complexity
| Structure
| Digital
Computers
| .
| .
| .
| Automata
| - to be developed
| .
| .
| |
Although the study of mathematics is very ancient, the objects that lead to the birth of the computer are very sparse until the early seventeenth century, when the craft generation starts. Various ways of using coins, beads, stones, and rope evolved. Among these the abacus and its derivatives are probably the most widespread.
DIGITAL CALCULA
|
SINGLE REGISTER - BEAD
The abacus is the earliest known computing device and the first hand-held calculator. It postdated the invention of the decimal system by the Egyptians circa 3000 BC. The Greeks and Romans built and used the abacus based on Hindu-Arabic numerals. Unlike earlier notations and devices using stones and marks, the abacus utilizes positional notation, including the representation of zeros, differences, with capabilities for multiplication and division. The Chinese abacus has beads in groups of 5 and 2, representing decimal digits. The Japanese first modified this to 5 and I and then 4 and I, a system known as bi-quinary representation that was also used in early electronic digital computers such as the IBM 650 (ca 1955). In the operation of the abacus, a single register machine, the moving of the beads also immediately provides the answer.
|
By 1620, the beginning of the craft generation, the abacus and counting table devices were in use and mathematical tables were made. In printing, the ability to use movable type was far ahead of paper technology, but a need existed for a convenient calculator or lookup table. John Napier of Herchiston, a mathematically oriented scholar, was bent on making long multiplication "free from slippery errors." His two major inventions logarithms (1614) and an inscribed set of rods or bones (1617) with number series that could be carried in the pocket and used as a look up table, immediately became quite popular. The bones were finely crafted sets that were sometimes paired with an abacus or a slate as a storage device. Although they are classified as manipulable tables, it can readily be seen, that their existence might have stimulated ideas for mechanical calculators. The invention of logarithms did, in fact, lead to the rapid development of slide rules, analog calculating devices. In 1620, Gunter placed the logarithmic scale on a rule and then a sector, and these devices rapidly came into widespread use satisfying the growing needs of exploration and trade. The speed of adoption of such devices, carried by navigators, was rapid, with developing trade and exploration and the ease in which they could be copied and crafted. Scientifically the use of logarithms and slide rules were aids to the development of mathematics and use of the mathematical tools in astronomy and for the academicians in the age of enlightenment. Thus, two devices, the bones and the development of rules with logarithmic scales, mark the beginning of the craft generation that was to last about 200 years.
Non-human interface is the first criteria that divides Memory. The earliest aids to human memory were neither machine writable or readable, ranging from stone markings, to beads, and papyrus scrolls. This group also includes the hand-crafted and personally read Napier's bones. The next Order of Memories are those that are either writable or readable by machine, ranging from printed books to semiconductor ROMs. And finally the third Order, both machine writable and readable did not begin to develop until the Electro-mechanical Generation.
NON-MECHANICAL MEMORY
|
FIXED PHYSICAL STATE
Napier's bones act as tables that can be rotated. Each
rod is inscribed with a set of numbers facilitating the
multiplication and division of large numbers. John
Napier, Laird of Merchiston in Scotland, invented the rods
and described them in his RABDOLOGIAE, (1617). He wrote
that the multiplication and division of great numbers is
troublesome, involving tedious expenditure of time, and
subject to "slippery errors." His tables reduced these
difficulties to simple addition and subtraction, and won
immediate recognition. A set of Napier's bones is usually
made of boxwood or ivory and often contained in a box or
case that would fit in a pocket. A set usually contains 10
rods, plus extras representing squares and cubes.
|
WRITABLE OR READABLE MEMORIES
PAPER - RANDOM ACCESS
|
Analog calculators work by analog, that is, they create a physical model of a mathematical problem. Many physical situations can yield mathematical results, provided they can be interpreted properly. The extent of a lateral or a rotational movement of a mechanism or the voltage level on a wire are examples of quantities which can be used to represent numbers. The most important breakthrough for analog calculators, however, came with the invention of logarithms by John Napier in 1614. This enabled the processes of multiplication and division to be carried out by addition and subtraction through proper positioning of number series along sliding rules. The results are interpolated between the marks on the rule. Other types of analog calculators include devices used in drafting, measuring and integrating, e.g., parallel, rules, planimeters, pantographs and harmonic analyzers.
The families in this order are divided according to the complexity of the mechanism itself single part, two-three part, multiple part, complex and programmable. This reflects a rough evolutionary development with multiple part devices not developing until mechanical tooling was improved, in the early nineteenth century.
SINGLE PART
DRAWING INSTRUMENTS
|
GUNTER RULE
About 1607 Edmund Gunter devised a scale that was to be
the predecessor of the modern slide rule. In 1623 he
published a description of this scale that is composed of two
scales of the logarithms from I to 10 placed end to end.
Although Napier conceived of the logarithm allowing
multiplication or division to be accomplished by addition or
subtraction, Napier relied on look up tables.
|
SECTOR
The sector is used to solve problems of proportion and
works on the principle of similar triangles. Sectors were
made with a variety of scales for use in calculation by
navigators, surveyors, gunners, and draughtsmen. At first
sight they look like a jointed rule usually made of ivory,
brass, wood, or sometimes silver. First described by both
Galileo in Italy and Thomas Hood in England the sector was in
use by 1600.
|
SLIDE RULE
In 1654, Robert Bissaker made the first real slide rule
in which the slide worked between parts of a fixed stock.
(Pugh 1975) The term slide rules applies to all instruments
designed so as to allow relative motion between the indices
and the scales. The classification used here is that
established in the Science Museum Catalogue i.e., straight,
circular, spiral or cylindrical, and log-log. The collection
illustrates the improvements in slide rules. Originally made
of boxwood, brass or ivory, in 1886 Dennert and Pape started
to use scales on strips of white celluloid to give much
greater distinction in reading. The spiral and cylindrical
scales allowed an increase of effective length, hence
accuracy, without equivalent increase in size. It also shows
the diversity and specialization that resulted for peculiar
needs at particular times.
STRAIGHT SLIDE RULES
|
CIRCULAR SLIDE RULES
|
SPIRAL SLIDE RULES
|
LOG-LOG SLIDE RULES
|
DMCAT1.3f
The second pre-computer generation started about 1810 and was brought about by the change from hand craft to mechanical technology. Two machines establish the beginning of the period: the Jacquard loom and :he planimeter. In the 1790 's Joseph Jacquard integrated a design )based on the ideas of Bouchon, deVaucauson, and Falcon, for an automatic harness controlled by punched cards connected to an endless roll that would mechanize fancy weaving. This was shown at an 'exhibition in Paris in 1801 and by 1812, ten thousand Jacquard loons were in operation in France alone. (Strandh, p. 195). The planimeter, the first instrument for directly measuring an area bounded by an irregular curve, appears to have been invented by the Bavarian engineer, J. M. Hermann in 1814. It was improved by Lamule in 1816, and constructed in 1817. (Pugh, 1975) With the need for surveying and recording land ownership, the planimeter rapidly came into widespread use.
In the mechanical generation, hand-crafted slide rules were spawned for a wide variety of uses; by revenuers to calculate tax on alcoholic leverages, lumbermen for cordage, printers for paper quantity, and traders for interest rates. (Turner, 1980) A company still exists in the North of England that makes specialized slide rules. Although the technology is based on a previous generation and two-three part analog calculators do not need mechanization, they were improved by industrialized forms of production.
The production of mechanical calculators did not start at the beginning of this generation. In 1820, Thomas of Colmar, an insurance agent, experimented with a four-function calculator, but it was not built or distributed until the 1850s. The real flowering of the mechanical calculators began in the last ten years of the century when Baldwin, Burroughs, and Felt were in business in the U.S., and Odhner had started his company in Russia.
2-3 PART
LEVEL REFERENCE
|
INTEGRATOR
MULTIPLE PART
LEVEL REFERENCE
DRAWING INSTRUMENTS
|
The Digital Order, Class Calcula has five families: single register, two register, three-four register, complex and programmable. The use of the abacus, a single register, manually built, portable calculator, has not been challenged until the development of equally portable and inexpensive electronic pocket calculators. Abacus-type machines have been unique because with a skilled, accurate operator, they could carry out diverse and complex functions, including long multiplication and division. They had the characteristic of all single-register machines, i.e., the only record of the operator's input was the current result on the single register. The dual calculator Sharp-Elsi Mate with both a soroban and a four-function electronic calculator was manufactured to preserve a culture, i.e., to teach children to use a soroban and not to use the calculator. If abacus-like machines are so extraordinary, why in fact were mechanical calculators ever invented? Probably, because of the likelihood of human error, and desire for simple aids with some kind of memory to check the human operator.
The Pascaline (1645) is the first of the mechanical, single register calculators. All machines stemming from this, to the Comptometer, utilized one's complement arithmetic for subtracting. two register calculators, developed in the late nineteenth "century, were characterized by using the keyboard as one register and using bi-directional wheels for direct subtraction.
Three and four register calculators were derived from Leibniz's concept of a stepped-wheel mechanism allowing an automatic carry, thus multiplication and division. Otto Steiger's millionaire, a heavy brass machine, based on purely mechanical principles, also had the first fully automatic multiply. Millionaire production came to a halt in the thirties, these machines were kludged with key punches and motors to meet the growing competition of electric motor-driven machines.
In the 1870s, both Frank Baldwin and Wilhelm Odhner developed a compressed version of the stepped wheel device with one large wheel and all operations based on its rotation. This type machine was widely distributed in Europe under the names Odhner and Brunsviga. its concept was most refined as the Curta, produced through the sixties.
In 1911 when Baldwin was old enough to retire he met Frank Monroe and they started the Monroe Calculator Company (Chase, 1980). The Monroematics, electric calculators, were among the first electrified automatic machines.
Four-function electronic calculators are with us, and school children and everyone needing to balance a check book have become about attached to them as they are to their watches. The inexpensive, the four-function electronic pocket calculator has replaced almost all other forms of analog and digital calculators. Complex digital calculators stem from Babbage's difference engine, built by Scheutz as a project.
SINGLE REGISTER
|
PASCAL WHEEL
KEYED WHEEL
One of the first trained operators, George W. Martin wrote Felt on November 6, 1887, "...in accordance with your request I have called on as many businessmen as I will have time to call on owing to the fact that the Gas Co. has written for me to come to work next Monday morning. The names and addresses are as follows: Sprague Warner and Co., Michigan Avenue and Randolph Pelkin and Brooks, Lake and State Streets, Melville E. Stone, Editor, of the Daily News, and the Freight Auditor of the C.B.&Q RR. These Gentlemen are very much please with the machine and say they will give it a trial as soon as you put it on the market." (Turck, 1921, p. 71) According to Turck, "significant proof of Felt's claim as the first inventor of the modern calculating machine is justified by the fact that no other multiple-order key-driven calculating machine was placed on the market prior to 1902." (Turck, 1921, P. 75)
|
|
STEPPED WHEEL
In 1820, Chevalier Charles X Thomas of Colmar designed and introduced the first multiplication machine made commercially available for general sale. Although it was not patented until 1851, the main features of the 1820 design remained unaltered. The mechanism has three parts, concerned with setting, counting, and recording respectively. Any number up to 999,999 may be set by moving the pointers to the numbers 0 to 9 engraved next to the six slots on the fixed cover plate. The movement of any of these pointers slides a small pinion with ten teeth along a square axle, underneath and to the left of which is a Leibniz stepped wheel. The Leibnix wheel, a cylinder having nine teeth of increasing length, is driven from the main shaft by means of a bevel wheel, and the small pinion is thus rotated by as many teeth as the cylinder bears in the plane corresponding to the digit set. This amount of rotation is transferred through one of a pair of bevel wheels, carried on a sleeve on the same axis, to the 'results' figure wheel on the back row on the hinged plate. This plate also carried the figure wheel recording the number of turns of the driving crank for each position of the hinged plate. The pair of bevel wheels is placed in proper gear by setting a lever at the top left-hand cover to either "Addition and Multiplication" or "Subtraction and Division". The "results" figure wheel is thereby rotated anti-clockwise or clockwise respectively.
AUTOMATIC STEPPED WHEEL
ROTARY
|
CARD-CONTROLLED - LOOM
|
The Origin of Punched card program control can be traced to 18th century developments in the French silk weaving industry.
|
COPIER
|
TYPEWRITER
|
The inventions that were critical for the electro-mechanical generation were fundamentally in place by 1900. These include the use of electro-magetics, electric-driven motors, battery-driven circuitry, and relays. Links and switches with telegraphy and telephony were developed throughout the mid-nineteenth century. Power for the early telegraphs was generated in conjunction with the railway system. Most early systems were point to point, along lines minus the technology for development as a network. The Hollerith tabulator and sorter developed for the 1890 census provides a truly significant project leading to a new generation. Its first commercial application was not until 1895, when a version was installed for accounting at the offices of the New York Central Railroad. (Randall 1973, p. 126) The 1900 census saw improvements in the system with the addition of automatic card handling mechanisms. In 1901 the first patent application for a motor-driven calculator was made. (No. 726,803 "The Universal Accountant" issued to Frank C. Rinche, April 28, 1903) The electric motor driven calculator was not produced in quantity until the 1920s. (Chase 1980)
Although the pieces of the technology were known prior to 1900, the infrastructure of the electricity grid had not been installed. This was essential to transform the to useful tools. On September 4, 1882, the first American power company, the New York Edison Illuminating Company, started generating electricity at the Pearl Street Station. (Stein, 1976, p. 244) Edison and others had difficulty raising money for these capital intensive projects and electrification had to be established as the infrastructure to support the use of electric-mechanical devices.
TWO REGISTER
KEYED WHEEL
|
3-4 REGISTER
MOTOR-DRIVEN WHEEL
|
COMPLEX
ACCOUNTING TABULATOR
Working as a statistician at the United States Census Bureau,
German-American Hermann Hollerith first conceived of using
punched cards as data carriers for the 1890 census. The 1880
census had taken over seven years to complete, the population
then numbering over 50 million and increasing rapidly.
|
WRITABLE OR READABLE
|
MECHANICALLY STABLE
RANDOM
|
TELEGRAPHY
NEEDLE TELEGRAPH
|
DMCAT 1.5
The mid-thirties brought needs for increasingly complex engineering calculations. George Stibitz recalled:
| In 1937, Bell Labs began to need greater calculating power for development in mathematical form as a theory of communications engineering. The basic principles were expressed in terms of complex numbers because they nicely represent the characteristics of alternating currents used by the power and communications industries. Twelve girls (if you don't mind the expression) did nothing but calculate complex numbers with 8 place precision using desk calculators. The arithmetic of complex numbers when it has been converted to multiple operations with real numbers and carried out on desk calculators is even more tedious and subject to errors than bookkeeping. Furthermore, the computing load was increasing rapidly. (Stibitz, 1980) |
| The human cum desk calculator (10 seconds per multiplication) would then spend about 2 hours on the multiplying; and with our estimate of a factor 6, about II hours doing an individual trajectory. This was a little right, perhaps a little low. The Harvard-IBM machine (3 seconds) required about 2 hours; the Bell machine (I second, about 2/3 hour; and the Mark II (0.4 seconds) about 1/4 hour. The differential analyzer took, as we have said, about 10-20 minutes. ... None of these was sufficient for Aberdeen's needs since a typical firing table require perhaps 2,000-4,000 trajectories assume 3,000. Thus, for example, the differential analyzer required perhaps 750 hours --30 days-- to do the trajectory calculations for a table. (Goldstine, 1972, p.138) |
The EDVAC report, written by Von Neumann, and based on the work of Eckert, Mauchly, Burks, and others involved with the ENIAC project, puts down the realistic specifications for the general purpose, stored program computer. (Von Neumann, 1945) It excited the academic community, and led to the origin of a number of computer projects. (Bigelow 1980) Similarly, a report by Alan Turing in England spawned interest there. A number of projects in laboratories and universities that developed between 1945 and 1950 then convinced the scientific, government and business communities of the reality of potential of the stored program, general purpose digital computer.
Although a driving meta-need was the war effort. World War II was over by the time Von Neumann was specifying the IAS project at Princeton. He identified a second meta-need, that of good weather prediction. The equations for greater accuracy in prediction were known in 1911-12 but time consuming to compute. Von Neumann's first experiments were so successful that as a result the U.S. set up a statistical weather prediction service. No conceptual breakthroughs had been made: it was only a case of carrying out the computations more carefully and with greater speed. With the advanced fourth generation computers, the one week theoretical limit of weather prediction as understood as a subset of celestial mechanics has not been reached. (Leith 1981)
COMPLEX
RELAY CALCULATOR
EQUATION-SOLVER
|
PROGRAMMABLE
RELAY CALCULATOR
The EDSAC
The ENIAC
Whirlwind Word length: 16 bits; Memory size: 2048 words; Speed: Approximately 42,000 single address instructions per second; Clock rate: I Mhz; 2 Mhz (for arithmetic element); Arithmetic element: Accumulator, A and B registers. Instruction format: Single address 5 bit op code and II bit address; Power consumption: Approximately 150,000 kw; Size: Occupied Barta Building, Cambridge. Component count: 5000 vacuum tubes and 11,000 crystal diodes; Availability: >95%; Maintainability: Used marginal checking of grid and screen bias voltage; Project leaders: Jay W. Forrester and Robert Everett. Project start: 1945; Operated: November, 1950 with 256 words; and August 1953 with core memory. Decommissioned: at MIT in May 1959; operated at Wolf R&D from 1963-1973; Moved to Digital, 1974. Use: Prototype for Air Defense Computer, precursor to IBM built AN/FSQ7 computer. Used to develop Linvill's sampled-data system theory. Achievements: First core memory. First high speed, parallel computer for real time. Control organized in an array permitting diodes to be used for specifying register transfer operations needed for designing each instruction in what Maurice Wilkes later described to be micro programmed. First use of marginal checking to detect weak components. Self checking procedure for faulty components. First use of cathode ray tubes for light pen input. Data transmission via phone lines; vacuum tube process improvements.
|
Industry itself and its leaders had been changed by the technological advances of the war period. Goldstine states:
In my opinion, it was Thomas Watson, Jr. who played the key role in moving IBM into the electronic computer field. When he came out of the Air Force in 1945 his experience as a pilot had apparently convinced him of the fundamental importance of electronics as a new and prime technology for our society. He therefore exerted considerable pressure on IBM..." (Goldstine, 1972, p. 329)
Word Length: 31 bits, including a sign bit, but excluding a blank spacer bit Memory Size: 4096 words Speed: .260 milliseconds access time between two adjacent physical words; access times between two adjacent addresses 2.340 milliseconds. Clock Rate: 120 Khz Power; 1500 Watts Arithmetic element: Three working registers: C the counter register, R the instruction register and A the accumulator register. Instruction format: Sixteen instructions using half-word format. Technology: 113 vacuum tubes and 1350 diodes. Number Produced: 320-490 First Delivery: September, 1956 Price: $47,000 Software: ACT I (Fortran type compiler) Successor: LGP-21 Achievements: With the Bendix G-15 the first of the desk-sized computers offering small scale scientific computing. Revolutionizing the computer industry with the potential for low-cost distributed processing.
COMPONENTS
LOGIC MODULE
|
WAVE STORAGE
CYCLIC
Mercury was used to propagate an acoustic wave and hold information. A two meter tube held about 1000 bits, with a delay time of approximately one millisecond with a bit separation of about one microsecond or two millimeters. Early computers such as the Pilot ACE, EDSAC, and Bureau of Standards computers used both long and short delay lines. |
ELECTRIC CHARGE
RANDOM
|
MAGNETIC FLUX
RANDOM
|
"What is a Transistor?
Transistors are made from silicon by the introduction of minute quantities of impurities that determine the electrical properties of the host material. By precisely controlling both the location and the concentration of impurities (called dopants), engineers can build up the transistor structure.
Doping impurities come in two types. The first adds free electrons to the silicon, converting it from a near insulator to a conductor of electricity (although the conductivity is much less than that of a metal). The second type removes electrons from the bonds keeping the silicon atoms in the solid, leaving behind electron vacancies or 'holes'. The holes behave like positively charged carriers of electricity and thus the second type of dopant also raises silicon's electrical conductivity. Silicon that conducts electricity by way of free electrons is called n-type, whereas material that conducts by the way of holes is called p-type.
Transistors consist of three segments of doped silicon back to back, as it were. The sequence of segments is important; the allowed orders are n-type-p-type-n-type and p-type-n-type-p-type. There are two general classes of transistors, but both can have either the n-p-n or p-n-p sequence of doped silicon segments. The historic first transistor built at Bell Laboratories in 1948 is called a bipolar transistor because electrical current flowing through the device from one end to the other passes through both n- and p-type silicon and both electrons and holes contribute to the current. Bipolar transistors are also called current controlled because a small electrical current entering the device through the center segment controls whether the device as a whole conducts electricity. A voltage applied only to the two end segments will not cause the transistor to conduct electricity. Viewing the transistor as a switch, one says that the current into the center segment turns the switch on or off.
The second class of transistor is the insulated gate field effect transistor. In this type of device, a thin insulating layer (usually silicon dioxide) is placed between the central segment and its electrode. A voltage applied to the electrode creates an electric field which converts the region of the central segment just under the electrode from one conductivity type to the other (n- to p-type of vice versa). Thus, field effect transistors differ from bipolar devices in two ways: they are actuated by a voltage applied to the central segment rather than by a current, and all the current is carried by one type of carrier in three segments of the same conductivity type.
With the invention of the integrated circuit in the late 1950s, it became clear that the field effect transistor offered distinct advantages because fewer processing steps were needed to make this type of device and because it took up less space in the silicon. The type of field effect transistor called metal-oxide-semiconductor (MOS) has become the dominant form of commercial integrated circuit. The biggest advantage of the bipolar device is switching speed. Thus, for those applications requiring this capability, such as high-speed logic circuits in computers, bipolar is widely used. Moreover, new forms of bipolar circuits that are more amenable to miniaturization than the older types are being investigated and may well turn out to be important as MOS microcircuits in the next generation of microelectronics, the VLSI era," (Robinson, 1980)
By 1960 transistors had replaced tubes as the technological base for computers. Their properties, lending themselves to automated design and manufacture no longer meant that the innovative machines would come from handcrafted projects in laboratories and universities, but from industrial research and development. The end of the fifties saw the last spurt of laboratory built machines: Lincoln Lab's TX-0 (Transistor Experimental project), MANIAC 2, Bell Labs Leprachaun and ILLIAC II. in 1959, a Siemens 2002 was delivered to the Technical University of Aachen. The same year IBM introduced their fully-transistorized 7030, the 7090, and the 1401. in 1960, the CDC 1604 and 160, and Digital Equipment Corporation's PDP-1, the IBM 1620, and the UNIVAC 1105 were announced. The full range of computers were then available for purchase: ranging from business to scientific, and from small to super, i.e., from $100,000 to $10,000,000.
The early sixties brought the space race creating new computing needs in science and education. This generated new demands for computing power that, once available, led the first generation of "hackers" to enhance the machines into super toys. The Gemini flight inspired a group in Cambridge to use the computer scope to simulate space flight and space wars. Active communication between users from coast to coast rapidly developed into a computer game culture. The children of the first hackers started to college in the eighties and are as distinctive as the so-called TV generation since they grew up with computers as playmates.
Simultaneously, business was beginning to define a need based on computing versus tabulating and sorting. Champine (80) has listed the phases that characterize the development of commercial applications. As early as 1955, the full range of business uses were envisioned that are continuing to create a need for larger and faster business computers. But he notes that only the leading edge users were implementing the intermediate level functions in the late fifties. Thus, business data processing only began to drive the development of computers with second generation machines.
LEVEL REFERENCE
3-4 REGISTER
Atlas Computer
IBM 7030 "THE STRETCH" IBM, 1961, Gift of Computer Service, Brigham Young University (D250.81).
Word Length: 64 bits plus 8 bits for parity and error checking Memory Size: I to 8 16k core memory stacks, self-contained each with its own clock, addressing circuits, data registers and checking circuits, addressing of up to 256k word locations. Data Transfer Rate: Addressing of memories and transfer of information from and to memories by a memory bus permits new addresses, information, or both to pass through the bus every 220 musec. Central Processor: The processor consists of the instruction unit, the look-ahead unit, a parallel arithmetic unit and a serial arithmetic unit. Multi-programming through program interruption and address monitoring, and overlapped or parallel execution of instructions is possible. Instruction Format: Halfword formats accommodate indexing and floating-point instructions. Fullword formats are used by variable-field-length instructions. Five instruction sets and 765 different types of instructions are used. Technology: Standard Modular System Transistor Cards. Used 150,000 high speed drift transistors, and provided interleaved magnetic core memory with 2.18 usec access cycle. Number Produced: 9 Price: $6-8 Million Project Start: 1954 Project Leaders: S.W. Dunwell; Gene Amdahl, John Backus, Werner Buchholz, B. O. Evans, Jerrier Haddad, Lloyd Hunter, Ralph Palmer, and John Sheldon First Delivery; April 1961 to the Los Alamos Scientific Laboratory Software: Algebraic and Fortran compiler Use: Large scale scientific research, for example: nuclear reactor design, hydrodynamic problems, problems in nuclear physics. Achievements: Techniques for parallel processing and multi-programming were interleaved memories, instruction look-up units, overlapping fetch and execute instructions, interrupt handling and address monitoring. The 7030 also introduced an 8-bit byte for character representation, up to 256 characters could be represented. The magnetic core memory developed for the STRETCH was also used on the IBM 7090.Innovations (adapted by Hurd, 1981)
|
LINC Computer, Lincoln Lab, 1961, (D118.79)
Word Length: 12 bits Memory Size: 2048 words Speed: Approximately 125,000 single address instructions per second Clock Rate: 500 khz using dec 4000 series modules Arithmetic Element: Six 12-bit registers Instruction Format: single and double operand, multi-mode Technology: Discrete transistor using dec 4000 series modules Power Consumption: 1000 watts Size: 69"x32"x32", plus separate tape, keyboard, console, and interconnection boxes. Price: $43,600 Project Leaders; Wesley Clark and Charles Molnar Project Start: 1961 First Shipment: March, 1962 Withdrawn: December, 1969 Number Built: 50 total, 21 by DEC Successors: DEC LINC, LINC-8, PDP-12 Achievements: Laboratory system to accept analog and digital inputs directly from experiments and to provide signals for control. First truly personal computer with automatic file system via two LINC tapes, interactive program editing, development and control via CRT. |
LINC-8, Digital Equipment Corp, 1965, (D119.80).
Word Length: 12 bits Speed: Approximately 667,000 memory accesses per second Clock Rate: I Mhz (same as PDP-8) Instruction Set Processor: Both LINC and PDP-8 Arithmetic Element: Four PDP-8, six LINC 12-bit registers Instruction Format: Single and double operand, multi-mode, 12 bit instructions Technology: DEC "Flip Chip" R-series general purpose modules. (Discrete components) Power consumption: 2,000 watts Size: 69"x32"x33" Price: $38,500 Project start: 1965 First shipment: August, 1966 Withdrawn: December, 1969 Predecessor: LINC Successor: PDP-12 Achievements: System where both processors could operate in parallel. Utilized either LINC or PDP-8 software. |
PDP-1, Digital Equipment Corp, 1960, Gift of Inforonics Corp (D116.79).
Word length: 18 bits Speed: 100,000 single address instructions per second Clock rate; 5 Mhz and 500 Khz for input-output Arithmetic element: Accumulator and input-output instruction format: Single address 5 bit op code, I indirect bit, 12 address bits. Extended field with 15 address bits. Technology: Early second generation Digital 1000 series 5 Mhz and 4000 series 500 Khz systems modules Power consumption: 2160 watts Size: 69"x88"x28" Price: $120,000 Project leader: Benjamin Gurley Project Start: Summer 1959 First Shipment: Bolt, Beranek and Newman, November 1960 Number built: 50 Achievements: First commercial computer with graphics display. Operation as time shared computer, BBN, September 1962. Original space war program by Steve Russell at MIT. |
PDP-7, Digital Equipment Corp, 1964, Gift of Computer Science Department, Worcester Polytechnic (D143.80).
PDP-8, Digital Equipment Corp,
Word length: 12 bits; Memory Size: 4096 words (expandable to 32,768 words); Speed: 333,333 signed address instructions/second; 1.5 microsecond memory cycle time; Clock rate: I Mhz; Arithmetic element: accumulator and 8 auto-index registers in memory; Instruction format: Single address 3 bit op code, indirect bit, I page bit and 7 page address; 32,768 word addressable memory; Technology: Digital R-series logic; Power consumption: 780 watts; Size: 8 cubic feet; Number produced: approximately 5,000; Price: $18,000 with 4096 word memory and teletype type 33ASR; Project start: 1964; First delivery: April 1965; Predecessor: PDP-5; Successors: PDP-8S, LINC-8, 8-1, 8-L, 8-F, 8/M, 8/A, VT78; Software: PAL-8 assembler. Macro 8 assembler, Fortran II, DDT (Symbolic debugger), Editor, RT-8 and OS-8 operating stand-alone operation systems using Dectape and diskpaks; Use: Real time control and data collection. First "OEM" computer. Data communication. Small business data processing. Timeshared computation for very low cost/terminal; Achievements: Originated concept of minicomputer; Provided the lowest cost computation and performance/cost at the time; Producible in high volume manufactured using wire-wrap technology; Improved ease of interfacing (first DEC computer to use I/O bus structure); By packaging, price and supply established the two tier supplier/OEM structure; Lowest cost per terminal with TSS/8 (smallest scale timesharing system). |
PDP-12, Digital Equipment Corp, 1967, (D156.80).
Word length: 12 bits Speed: Approximately 667,000 memory-processor accesses per second Clock rate: I Mhz (same as PDP-8) Instruction Set Processor: Both LINC and PDP-8 Arithmetic element: Four PDP-8, six LINC 12-bit registers Instruction Format: Single and double operand, multi-mode 12-bit instructions Technology; DEC "Flip-Chip" general purpose modules. Discrete components. Power Consumption: Less than 2000 watts Size: 76"x35"x33" Price: $28,000 Project Start: June, 1967 First Shipment: June, 1969 Withdrawn: June, 1975 Number built: 1,000 Predecessors: LINC, LINC-8 Achievements: Improved price, price per performance and larger display. Lowered LINC-8 cost by building a single physical processor to execute either LINC or PDP-8 instruction set. |
TX-0 Computer, Lincoln Lab, 1956, (D154.75).
Word Length: 18 bits Memory Size: 8192 words Speed: 80,000 single address instructions per second Clock Rate; Variable, controlled by delay-line (max rate = 5 Mhz) Arithmetic Element: Accumulator; In-Out Register for program-controlled Input-Output; Index Register Instruction Format: Five bit op code, (2 bits initially used) + 13 bit address (16 bits for initial 65,536 word memory) Technology: Discrete transistor circuits and core memory Power consumption: Approximately 5,400 watts Air Conditioning: 15 tons Size: Built into 9000 square foot room at MIT Component Count: 3,600 surface-barrier transistors (SBT) of Philco type 2N240 Total Hours: Approximately 50,000 hours with 12 transistor failures Project Staffing: Lincoln Laboratory Division 6, Group 63; William Papian, head; Wesley Clark, logical design; Kenneth Olsen, circuit design and construction (followed by Benjamin Gurley) Richard Best and Jack Mitchell, memory design. John Clarke supervised construction. Project start: Late 1955 Use: Research on electro-physiological signal processing; speech analysis and synthesis; picture processing; simulation of sensory aids for the blind; bubble chamber photograph analysis; handwriting analysis; interactive programming; symbolic program tracing and debugging. Achievements: Tested transistorized circuitry for use in computers. Tested a large, 65,536 word (18 bit + parity bit per word) vacuum tube driven core memory. Improved real-time interfacing. |
"Tomorrow: The Thinking Machine", CBS, 1961, B&W, 3/4" videotape,
Running time: I hr. (V6.81) Artificial intelligence is the topic of "The Thinking Machine," a 1961 episode of the CBS News Tomorrow show, narrated by Jerome Weisner and David Wayne. Machine "learning" is compared with human and animal behavior. Highlights include an interview with Claude Shannon, a robot-sequence clip from the silent film classic "Metropolis", and three versions of a TV western written on MIT's TX-0 computer. |
MAGNETIC
RANDOM, Rope
|
WAVE STORAGE, CYCLIC
|
Magneto-strictive Delay-line stores hold information as a series of impulses circulating continuously along a closed path. In a magnetostrictive delay-line electrical impulses signifying data are converted into stress waves which travel the length of the nickel wire. The application of a magnetic field to the wire causes it to change dimension thus converting electrical impulses to stress waves, or vice versa. Coils similar to those found in an electro-magnet are used for inserting and recovering digital information from the delay-line. The Elliott Brothers' Computers in England were the first to use the magnetostrictive principle for storage of data. (Lavington, 1980)
|
MAGNETIC FLUX
|
RANDOM, CORE
Cores are made of ferromagnetic material that is able to become strongly magnetic when subjected to relatively weak magnetic forces. A magnetic field is generated in the vicinity of any conductor that is carrying a current. The direction of the magnetic field is related to the direction of the current flow in such a way that reversing the direction of the current results in a reversal of the direction of the induced magnetic forces. Each core has four wires: two which write selecting the proper one in a co-incident (x-y) axis. A third wire reads and a fourth wire inhibits a build up of energy. A number of core planes are then piled into a core stack or cube and in the transistor and integrated circuit computer generations were the most prevalent type of primary memory.
DISK
PLATED WIRE
|
??COMPONENTS
LOGIC MODULES
|
DMCAT 2.3
DIGITAL CALCULA
FIVE OR MORE REGISTER
|
DIGITAL COMPUTER
|
COMPONENTS
SECONDARY MEMORY
|
DHCAT2. 4
DIGITAL COMPUTER
VAX Computer Exhibit
MINC, (Modular Instrument Computer) Digital Equipment Corp. 1975, (D155.80). Word length: 16 bits Memory size: 32,768 words Speed: Approximately 200,000 single instructions per second Clock rate: 3 Mhz Instruction set Processor: PDP-11 (LSI-11}) Arithmetic element: Data path on an LSI chip, 8 general purpose registers Instruction format: Double operand, multi- mode, 16 bit instructions Power consumption: Approximately 500 watts Size: Roll around cart (24"x30"x40") Component Count: 4 LSI chips forming the LSI-11 processor. 300 MSI and LSI chips for memory and peripherals Project start: August, 1975 Packaging model demonstrated: August, 1976 Running system demonstrated: August, 1977 Product announcement: October, 1978 First shipment: December, 1978 Number produced: 1500 annually Input-Output: Real-time plug-in modules for analog, digital event processing and signal conditions. Graphics CRT. IEEE 488 and serial communications lines. Software: Real-time and graphics BASIC. Optional languages and facilities available on PDP-11. Use: Science-based discipline computation, including general purpose programming, mathematical modeling, graph plotting, laboratory management. Real-time use including data acquisition, signal processing or experiment control. Achievements: Improved human interface as scientific and laboratory computer through software, modular hardware and documentation. Improved cost and performance per cost of ownership by portability, higher mean time between failures (MTBF). customer installation, built-in service and diagnostics, and direct phone link to factory for information.
PDP11/23 Micro-computer Processor Module, Digital Equipment Corp. 1979. (D33.80). |
COMPONENTS
LOGIC MODULES
|
DMCAT2.5
If you have comments or suggestions, Send e-mail to Ed Thelen
Go to Antique Computer home page, Go back to Catalogs
Originated June 8, 2000
Updated April 22, 2013