Switches at the heart of every computer
Whatever else a computer does, its circuits are built from switches that can be in one of two states, on or off. The history of early hardware is largely the history of better switches. The first practical ones were mechanical or electromechanical, and the next were electronic. Understanding them makes it easier to see why computing changed so sharply in the 1940s and 1950s.
Relays: switches with moving parts
A relay is a switch operated by an electromagnet. When current flows through a coil, it pulls a metal contact closed. Relays were already widely used in telephone exchanges and other equipment, so they were an obvious building block for early calculating machines.
In the 1930s and 1940s, relay-based computers were built by several groups, including Konrad Zuse in Germany, researchers at Bell Labs in the United States, and a team behind the Harvard Mark I, which was electromechanical and completed in the 1940s. Their reliability was reasonable, and they could run long calculations without human intervention. Their limitation was speed. Because a relay is a physical part that has to move, it can only change state so quickly, and the contacts wear or collect dirt over time.
Vacuum tubes: switching without motion
A vacuum tube is a sealed glass bulb in which a heated element releases electrons that flow to another electrode. A small voltage applied to a third element, called a grid, can control that flow. The result is a device that can amplify a signal or act as an on-off switch with no moving parts.
Because electrons move much faster than metal contacts, tube circuits could operate far more quickly than relays. Researchers experimented with tubes for computation in the late 1930s and during the Second World War. Machines such as Colossus in Britain, used for codebreaking, and the ENIAC in the United States are well-known examples, and the development of electronic computers more broadly owes much to such wartime work. Credit for the “first” electronic computer is shared and debated among several projects, and it depends partly on definitions.
What it was like to use them
Tube-based machines were remarkable but demanding.
- Size and power. A machine could contain thousands of tubes and fill a large room. Each tube needed power for its heated element, so electricity use was high.
- Heat. All that heating required serious cooling. Rooms could become uncomfortably warm, and the heat itself shortened component life.
- Reliability. Tubes burned out, in a way similar to light bulbs. With thousands in use, failures were a regular event, and finding a faulty one could take time. Operators and engineers developed routines for testing and swapping tubes.
- Wiring. Early machines were often programmed or reconfigured by physically changing cables and switches, though the idea of the stored-program concept, in which a machine holds its instructions in memory, soon changed this.
Some accounts of the period describe long stretches in which machines worked correctly, and others describe frequent downtime. Experience varied by machine, by maintenance practice and by how much the machines were pushed, so sweeping claims should be treated with caution.
Making memory with the same parts
Memory was a major challenge. Early memory systems included delay lines, in which signals circulated through a medium, and electronic storage tubes that held patterns of charge on a screen. Some machines also used magnetic drums. Each had limits in capacity, speed or reliability. Later, magnetic core memory, a grid of tiny magnetized rings, became a widespread solution in the 1950s and was used for years. Punched cards and paper tape remained the usual way to enter data.
Enter the transistor
The transistor was demonstrated at Bell Labs in the late 1940s. It is a solid-state device, which means it has no heated element and no vacuum, and it can act as a switch or amplifier. Compared with tubes, transistors are generally smaller, use less power, give off less heat and tend to last longer.
Early transistors were difficult to manufacture consistently and cost more, so the shift was gradual. During the 1950s, experimental and then commercial machines began using them, and by the early 1960s transistor designs dominated new computers. The move to transistors made machines more dependable and reduced the effort spent on maintenance. It also made computers practical for more organizations, helping the mainframe business grow.
Did tubes disappear entirely?
For computing, tubes were largely replaced, but they did not vanish from all electronics. They remained in use in some radio transmitters, certain audio equipment and specialized devices, and the display tubes that showed early computer output are a related but different technology. Enthusiasts also keep old tube machines running, which says something about both the craftsmanship and the effort involved.
Why it matters
The jump from relays to tubes to transistors illustrates a pattern seen throughout computing: a component that is faster, smaller and more reliable than its predecessor opens up new uses, and those uses create demand for further improvement. The next major step, putting many transistors and their wiring onto one chip, built on exactly this foundation.