What engineering advances paved the way for "modern" computers?

Question

By "modern computers," I mean electronic programmable computers such as those that were developed about the middle of the 20th century.

My understanding is that early computers such as Charles Babbage's "calculating machines" worked mainly on mechanical principles, like a modern "abacus". Apparently later machines were more electronically based. I'm talking about computers that evolved from using vacuum tubes, to transistors to integrated circuits, and ultimately silicon chips.

What engineering advances enabled the above transition to occur from physical means (e.g. vacuum tubes) to electronic means (e.g. transistors) when it did (middle 20th century)?

Answer 1

The transistor

The transistor was the revolutionary replacement of the vacuum tube, which had been at the heart of computers for the first half of the 20th century. Vacuum tubes themselves had only two main problems: They were power-hungry and they were relatively big. Relative to their replacements, that is. They also had a tendency to burn out or leak during operation, which could prove disastrous.

In 1947, John Bardeen and Walter Brattain, along with William Shockley, their boss at Bell Labs, successfully amplified an electric current, using germanium. This "point-contact transistor", as it was called, was soon used to speed up computing and to make computers smaller and more efficient.

A good example of the transistor and the vacuum tube is the construction of the Manchester Computers, developed at the University of Manchester. The first, the Small-Scale Experimental Machine (SSEM) (developed in 1947), was a state-of-the-art testbed for new innovations in computing, such as the Williams tube. But it still used vacuum tubes. It had 550 valves and took in 3500 watts of power.

The SSEM's successor, the Manchester Mark 1 (developed in 1949), was much more powerful. It used 4050 valves and consumed 25000 watts of power. Yet the aptly-named Transistor Computer, built in 1955, used only 200 transistors and 1,300 diodes, and only used 150 watts. It wasn't the first computer to exclusively use vacuum tubes, but it was a huge step forward.

It's tough to say exactly why the transistor was created when it was (I'm answering the last part of your question now), but it could be argued that the computational advances of World War II (such as the Harvard Mark I) ensured that many new advances would be made in computing; the transistor happened to be one of them.

The integrated circuit

The integrated circuit, developed about a decade after the transistor, also had profound effects on computing. It was developed by Jack Kilby in 1958 - though many others were involved along the way, and there are disputes as to who should get the credit for inventing it first - at Texas Instruments. He used semiconductors to create an entire computer chip - the integrated circuit.

An integrated circuit can contain incredible amounts of transistors, and it is this complexity and compactness that make it so useful. Manufacturing was also much easier and quicker. Integrated circuits started a second computing revolution, which laid the groundwork for cheaper computers that could be available to the masses.

---

Now that the question is focused on the transition between vacuum tubes and transistors, I'd like to add something about semiconductors, because they play a key role in transistors.

Semiconductors allow for good conduction of electricity, but one of their really useful properties is that their conduction can be modified in a process called semiconductor doping. This adds "impurities" to the semiconductor, adding either electrons or holes. Semiconductors can be either n-type or p-type - n-type semiconductors have an excess of electrons, while p-type semiconductors have an excess of holes. These can be combined to form a diode.

Another relevant development was the creation of the Czochralski process, which allows crystals to be grown for semiconductors. This also involves semiconductor doping, and has allowed for semiconductors to be produced on a large scale, making it easier to build transistors.

---

Are there other technologies that have been crucial to computer development? Of course. I could cite the vacuum tube, cathode ray tube, solid-state drives, and a host of others as crucial to computer development. But the transistor and the integrated circuit were the two main players in the development of the "modern computer" in the relevant time period here - the mid-20th century. I suppose you could make cases for other technologies, but I'd certainly rank them at the top of the list.

Answer 2

I'm still confused about what you mean by "engineering advances". To me, things like vacuum tubes, transistors and ICs are technologies, and engineering is all about applying those technologies to real-world problems, such as building computers.

Engineers have typically taken each available technology to its practical limit — to the point where other issues become predominant. Examples of this would include:

• The ability to machine the individual parts of Babbage's mill to sufficient precision so that it worked accurately and reliably. This became the limit on how complex a mechanical calculator could be.
• Making vacuum tubes reliable enough that you could operate tens of thousands of them for at least a few hours before one of them burned out. This became the limit on how complex a vacuum tube computer could be.

Are these the sorts of engineering issues you're looking for?

With integrated circuits, the current mainstream technology, the engineering issues include things like minimum feature size, static and dynamic power dissipation, and defect densities. It's a complex problem to find the best combination of parameters that results in the most cost-effective (and profitable!) chips.

How did the discovery of vacuum tubes, transistors, and integrated circuits allow people to "engineer" computers in a way they hadn't done so before?

As I said, the physical properties of mechanical parts and the ability to machine them precisely limited both the speed and the complexity of mechanical and electromechanical (relay-based) computers.

The ability to mass-produce reliable vacuum tubes made it possible to build purely electronic circuits that matched (and eventually exceeded) the complexity of relay-based logic, while operating at much higher speeds: tens and hundredss of kHz, rather than tens of cycles per second or less. However, the reliability of the tube filaments eventually became the dominant issue, once computers got to the point of needing tens of thousands of them.

The development of the transistor allowed circuits to be shrunk by at least an order of magnitude, but more importantly, the MTBF of transistors exceeded that of tubes by several orders of magnitude. Both of these advantages enabled the development of more complex computers. But the physical size of the transistor in its package still presented a problem in terms of the size of the computer, and the large number of discrete connections still presented a reliability issue.

The integrated circuit has solved both of those issues, with millions of transistors crammed into the volume that a single discrete transistor requires, and the connections being made by patterning metal directly on the IC surface. Both the complexity and the speed of computers has grown by leaps and bounds as we learned to make ever-finer patterns on the surface of the chips. We are only just beginning to see some fundamental limits on how far we can push this before we need a new fundamental technology.

Answer 3

HDE 226868 already mentioned the transistor in detail, so I will add many, many technological advances and theoretical backgrounds for the evolution of the computer.

• Metallurgy
  For the mass-market you need cheap and reliable methods to produce the desired electronic elements. For mass production you need integrated circuits and that means, especially if you go down to small dimensions, you need homogenous high-quality base material. Only with high-grade silicon wafers this was possible and while the Czochralski process from 1916 is cheap it did not give such good results for silicon like zone melting which was invented 1950/51.

• Boolsche Algebra und Cybernetics. The logical algebra developed by Boole allowed to build computers on binary states. If you ever had the pleasure to try using decimal systems on an electronic machine (BCD), you see how unbelievably easier it is to implement logical and numerical functions. With progressing complexity, it was also necessary to invent the necessary background for developing and controlling complex electronic system, so cybernetics as a new study began to develop. It really cannot be underestimated what Babbage and later von Neumann did: Instead of building specialized models for tasks they invented the idea of a programmable machine: Provide simple building blocks of commands and express your solution with this simple commands instead of rewiring the machine (Perhaps there are still some people here who literally "reprogrammed" computers by rewiring parts !). Allow conditional execution. Treat data and program as unit. The general-purpose computer was born.

• Photography and Chemistry Only photography was able to provide people to allow scaling down developed solutions cheaply, robustly and efficiently. Without photo lithography and the development of chemicals allowing the etching of microscopic structures, computers could be small, but they would be extremely expensive.