Page 1 of 2
Gordon Bell is responsible for many things, but the design of the most successful range of minicomputers, the PDP range, is probably the thing he is best known for. This is a story about when computers were big, but getting smaller...
Mainframes, especially the really big-as-a-house valve-based designs, have the attraction of heavy engineering and being first.
The latest microdesigns also have the attraction of power in a small and elegant package.
This leaves the mini-computers, the forerunners of the micro, slightly out in the cold and unloved.
The age of the mini-computer is almost completely ignored in the standard accounts of the development of computing. It is as if the world made the jump from mainframes to micros with only a Xerox workstation in between!
Yet it is in the design of the mini-computer we first see some of the features that we regard as essential in a modern computer. It is also the mini-computer that started the trend of bringing computing to the masses. Well, if not the masses, at least they brought computing out of the purpose-built computer rooms to a normal working environment.
Just as we tend to ignore the importance of the mini-computer we also ignore the people who built them.
Perhaps the most remarkable of these early pioneers is Gordon Bell.
Chester Gordon Bell
Born August 19th, 1934 in Kirksville, Missouri, USA
Early interest in electronics
Gordon Bell was interested in electronics from an early age. His father ran an electrical store in small town in Missouri and he helped out with small jobs such as putting plugs on - eventually graduating to rewiring houses!
As an undergraduate at MIT he studied electrical engineering and worked at a number of large electrical companies as part of his course. However, it was the newly emerging field of electronics that interested him the most. He tinkered with valves as a hobby - building an audio amplifier and took the two modules on digital techniques that his course offered. Back then digital electronics was thought of as not particularly mainstream or useful. Most electronics was analog.
In 1957 Bell won a Fulbright scholarship and went off to Australia to study at the University of New South Wales. This gave him a chance to put off the difficult question of what to do for a living. He didn't want to work for one of the big companies. He wanted an opportunity to work creatively on projects. Today he would almost certainly have started a software or hardware company in his garage - but this means of escape hadn't then been invented!
His time was spent writing software and giving digital design courses - the first at the university. He also met a fellow Fulbright student - Gwen - who on their return to the USA became his wife.
What to do next?
The choice presented to him was to work for Philco or to return to MIT to complete a doctorate. He opted to put off joining a company and built a statistical sound analysis meter at the MIT Speech Computation Laboratory writing the first Analysis by Synthesis program. It was suggested that he might like to work with the TX-0, the first in a line of mini-computers designed at MIT.
Following on from the sound analysis meter Bell decided to add speech analysis and recognition hardware and software to the TX-0. The task was next to impossible given the state-of-the-art at the time. He realised that it would take tens of years to achieve something that worked and this gave it the status of pure research rather than the engineering that he liked.
DEC and the PDP-1
As luck would have it there were engineering challenges a plenty right on his doorstep. Kenneth Olsen, another MIT graduate, had started a small company called DEC, standing for Digital Equipment Corporation, to build modules to work with the TX-0.
Bell had used some of these modules and knew he TX-0 well and DEC's products. In 1960 he joined DEC to do - just about everything - programming, architecture and detailed digital design. In the 60s "full stack" really did mean everything from the bare metal to the high-level language.
DEC's aim in life was to build smaller more affordable computers and this is exactly what they did with the PDP-1 - one of the first commercial machines to use transistors.
So daring was this approach to computing that DEC didn't want to risk ridicule by calling their new machine a computer - instead they called it a Program Data Processor, hence PDP.
It was big for a mini-computer - four 6 foot tall cabinets - and by our standards it wasn't very powerful and it wasn't very cheap. Even so at $120,000 it was a good choice as a laboratory machines compared to the alternative of renting a larger machine from IBM.
Sales of the PDP-1 were slow at first, but eventually picked up enough for DEC to build further machines. One stroke of luck was its use by ITT as a communications switch. Bell worked on the special hardware needed and is credited with the design of the first UART - Universal Asynchronous Receiver Transmitter. The UART is the basic component that turns the parallel data into a serial bit stream and, more difficult, turns the serial bit stream back into parallel data.
The next major design work that Bell undertook was the PDP-4 which was completed in 1962. The design was similar to the PDP-1 but it was simpler and offered nearly the same processing power for half the price.
The PDP-5 was designed as a front end data gathering computer for use with the PDP-4. The PDP-5 was at the time the worlds smallest and cheapest computer. The PDP-6 was a step in the other direction - a large multiprocessor machine. It wasn't a huge success, only 20 were built, but it was the start of Bell's interest in multiprocessor and large parallel machines.
The real breakthrough into the mass market was the PDP-8.
The new machine was small, a single cabinet, fast and cheap. It used a 12-bit word and small 4K-word core memory. At last transistors were cheap enough to allow the CPU to use fast registers rather than memory locations.
This was a computer which offered enough power for many jobs at only $18,000. It was an instant success. Over its life time 100,000 PDP-8s were sold. There was even a PDP-8 microprocessor which might have altered the course of PC history if DEC had decided to sell it for a low price to other manufacturers.
Gordon Bell designed the PDP-8's hardware and much of the software - it was his machine. After having built the best small computer available he opted not to build the PDP-9 but to take a change as associate professor at Carnegie-Mellon University.
He remained there from 1966 to 1972 working mostly on highly parallel multiprocessor machines such as the 16-processor C.mpp and the 50-processor Cm*.