SCIENCE
Meetings chaired by Andy Pepperdine unless stated

Phil Cooke, Member, on 26 September 2003

The lecturer started with a brief discussion on the title of the lecture. He explained that the main subject, the history of computers, came about because he had been privileged to spend the early part of his career working with many of the early pioneers. But the strap line, of "and how we depend on them" was there to enable the talk to have contemporary significance.

The early history of computers was then discussed. The first stored program computer ran at The University of Manchester in 1948. Prior to that there a number of specialised computing engines had been produced, but the Manchester machine is recognised as the first general purpose computer. It started out as a test system for a memory system invented by FC Williams, and based on a CRT (Cathode Ray Tube). Williams, Tom Kilburn the computer’s designer, and many of the early pioneers had worked at the Radar Research Establishment in Malvern during the Second World War. It was the Williams memory system that enabled the Manchester team to produce the world’s first stored programme computer, beating their much better funded competitors in the USA.

In parallel, Maurice Wilkes led a rival team at The University of Cambridge, with Bill Renwick, as Chief Engineer. EDSAC ran its first program in May of 1948. Like all the early computers EDSAC was a serial machine. This fitted in well with the memory technology, which for EDSAC was a number of mercury columns used as acoustic delay lines.

Before EDSAC was commissioned J. Lyons, of Corner House fame, had become fascinated with the potential for speeding up clerical work. This interest led to LEO, the world’s first commercial computer.

In the USA several well-funded teams were also at work. Most famous of the early pioneers is Von Neuman, a mathematician and theoretician. His most enduring contribution is the way in which a single memory system is used to hold both the programme and the data. Modern computers still use this architecture.

American designed computers quickly followed, but there also came a fundamental innovation: core stores. The inventor was Jan Forrester of Massachusetts Institute of Technology, MIT. The concept was to use a small torroid of a ferrite material with a square hysterisis loop. This technology enabled the design of reliable random access memory systems, with access times measured in microseconds. Core stores gave considerable impetus to the now emerging computer industry.

In Manchester, the commercial connection of Cambridge and Lyons was mirrored with collaborations between Ferranti and AEI/Metropoliton-Vickers, both then large electrical engineering companies. These links led to the production of the MV950, the first world’s commercial computer to use transistors of the long forgotten point contact variety. The lecturer explained that he made contributions to this computer during his first job after graduating. He went on to design the arithmetic unit for the AEI 1010, a first generation parallel computer that used early transistors and a core store. The lecturer showed a circuit diagram of a single register for this computer: it required about 30 discrete components, four power supplies, and took more than 20 square inches of printed circuit. All this for what is now classed as one bit of memory.

Meanwhile, the Manchester team were making plans to design a world beating supercomputer called ATLAS. In the USA the rivals IBM were designing STRETCH. The latter had massive design teams and Defence funding but it was the small and tightly knit group under Tom Kilburn that are remembered. ATLAS, like Concorde was a technical triumph, but not a commercial success. However pioneering concepts such as an operating system, virtual memory and large numbers of index registers are concepts that are now mainstream.

The author then outlined his personal contribution to ATLAS. He designed two memory systems, and both were based on magnetic cores. The main-store was assembled from units of 200kbits each occupying a double rack, with a speed of less than 1Mhz. The other was very specialised, and probably the only two-core-per-bit store to achieve commercial success. It enabled ATLAS to have an array of 128 index registers, each of 64 bits, and with an access time of 1microsecond. Compared to contemporary computers the specifications sound puny, but when ATLAS went live it doubled the total computing capacity in the UK.

During the period in which ATLAS was being designed electronics was changing rapidly. Transistors were becoming faster and cheaper, and electronics was being applied in many new areas. Moreover there had also been a new invention: the "Silicon Chip". By the time ATLAS was in production it was possible to produce a single storage bit on a single chip, and the lecturer told how he then moved into that side of electronics. By the mid 60s the technology enabled his then employer, Plessey, to create a 4 bit store on a single chip. The future of core stores was already looking problematic.

By the end of the 60s mainframe computers were commonplace, and mini-computers were driving innovation in industrial automation and data communications. There then followed a period of rapid change, a sort of virtuous circle. As affordable computers became available the time consuming and error prone tasks involved in designing and producing chips became automated, and new design tools became available. This enabled more complex chips to be designed, in turn leading to more powerful computers, each generation more powerful, smaller in size, and more affordable.

By the late 1960s silicon chips were displacing cores as the computer memory of choice. Then in the early 70s Intel launched the first microprocessor. It was a 4 bit computer called the 4004. Almost as soon as these chips were released the electronics industry saw the potential for innovation. The demand for chips grew exponentially and drove forward both the semiconductor and electronics industry. Hobbyists were also quick to see the potential for producing affordable computers, but the mainstream computer companies remained amazingly aloof.

Whilst IBM with their mainframes and DEC with their mini-computers grew and gained global significance, the semiconductor industry harnessed computers and intelligent instruments to design and manufacture chips of increasing complexity. Every couple of years chip complexity doubled. This rate of progress was famously forecast by Gordon Moore, an INTEL founder, and is known as "Moore’s Law" Figure 1 shows the result.

By 1980 chips were being produced with almost 100,000 transistors, enough for 10s of kbits of computer memory, or the 8086 microprocessor. Eventually IBM took notice of the potential offered by microprocessors, and used the Intel 8086 processor as the basis for the IBM PC. They made two decisions that were to change the face of computing. Firstly they decided to open-up the architecture so that third parties could produce hardware and software add-ons: and secondly they entrusted the supply of the operating system to a young entrepreneur called Bill Gates, and un-wittingly gave birth to Microsoft.

The lecturer then put forward the view that there are now few aspects of business or everyday life that have not been touched by the availability of cheap computing. He observed that whilst the effect of personal computers is almost self evident, the extent to which society has come to rely on computers hidden from view – what the trade calls "embedded"– is less well appreciated. He devoted the remainder of the lecture to this theme.

As a starting point he posed the question "In what way has life changed over the last 50 years?" This resulted in a list comprising:

 

• Ease of Travel

 

• Ease of communication

 

• Automation in the home

 

• Automation in the office

 

• Automation in manufacture

 

• Choice in the shops

He had then examined how one small part of one item, mass air travel, would not have happened without cheap computing. He started by listing ease of booking, well-advertised, reliable and safe, suitable aircraft, and fuel as non-technical pre-requisites for mass air travel. He then showed that these aspects depended on one or more of the following, hi-tech printing, cheap data/telecoms, DTP, big databases, TV advertising, computer aided design, smart machine tools, air traffic control, and sophisticated oil exploration and manufacturing technologies. He then showed how these were in turn reliant on more basic technologies. He listed electronic fonts, electronic formats for data, colour processing, graphics processing, mechatronics, graphics output, digital control, and signal processing. Each of the last list he explained was an enabling technology in its own right. The relationships are shown more clearly in Figure 2.

The lecturer concluded by observing that whilst everyone was aware that the world had changed vastly over the last fifty years, few were aware of how little it might have changed had not computers become ubiquitous. But this required computers to become cheap and small, which required silicon chips. However complex, chips would not have been possible without cheap and powerful computing for CAD, precision machines and intelligent instruments. So one depends on the other, and society now depends on both.

Phil Cooke