'tunny' Episodes

Of Heath Robinson Contraptions And The Colossus

     12/14/2021

The Industrial Revolution gave us the rise of factories all over the world in the 1800s. Life was moving faster and we were engineering complex solutions to mass produce items. And many expanded from there to engineer complex solutions for simple problems. Cartoonist Heath Robinson harnessed the reaction from normal humans to this changing world in the forms of cartoons and illustrations of elaborate machines meant to accomplish simple tasks.

These became known as “Heath Robinson contraptions” and were a reaction to the changing and increasingly complicated world order as much as anything. Just think of the rapidly evolving financial markets as one sign of the times! Following World War I, other cartoonists made similar cartoons. Like Rube Goldberg, giving us the concept of Rube Goldberg machines in the US.

And the very idea of breaking down simple operations into Boolean logic from those who didn’t understand the “why” would have seemed preposterous. I mean a wheel with 60 teeth or a complex series of switches and relays to achieve the same result? And yet with flip-flop circuits one would be able to process infinitely faster than it would take that wheel to turn with any semblance of precision. The Industrial Revolution of our data was to come.

And yet we were coming to a place in the world where we were just waking up to the reality of moving from analog to digital as Robinson passed away in 1944 with a series of electromechanical computers named after Robinson and then The Colossus. These came just one year after Claude Shannon and Alan Turing, two giants in the early history of computers, met at Bell Labs.

And a huge step in that transition was a paper by Alan Turing in 1936 called "On Computable Numbers with an Application to the Entscheidungsproblem.” This would become the basis for a programmable computing machine concept and so before the war, Alan Turing had published papers about the computability of problems using what we now call a Turing machine - or recipes. Some of the work on that paper was inspired by Max Newman, who helped Turing go off to Princeton to work on all the maths, where Turing would get a PhD in 1938. He returned home and started working part-time at the Government Code and Cypher school during the pre-war buildup.

Hitler invaded Poland the next year, sparking World War II. The Poles had gotten pretty good with codebreaking, being situated right between world powers Germany and Russia and their ability to see troop movements through decrypted communications was one way they were able to keep forces in optimal locations. And yet the Germans got in there. The Germans had built a machine called the Enigma that also allowed their Navy to encrypt communications. Unable to track their movements, Allied forces were playing a cat and mouse game and not doing very well at it. Turing came up with a new way of decrypting the messages and that went into a new version of the Polish Bomba.

Later that year, the UK declared war on Germany. Turing’s work resulted in a lot of other advances in cryptanalysis throughout the war. But he also brought home the idea of an electromechanical machine to break those codes - almost as though he’d written a paper on building machines to do such things years before.

The Germans had given away a key to decrypt communications accidentally in 1941 and the codebreakers at Bletchley Park got to work on breaking the machines that used the Lorenz Cipher in new and interesting ways. The work had reduced the amount of losses - but they needed more people. It was time intensive to go through the possible wheel positions or guess at them, and every week meant lives lost. Or they needed more automation of people tasks… So they looked to automate the process.

Turing and the others wrote to Churchill directly. Churchill started his memo to General Ismay with “ACTION THIS DAY” and so they were able to get more bombes up and running. Bill Tutte and the codebreakers worked out the logic to process the work done by hand. The same number of codebreakers were able to a ton more work. The first pass was a device with uniselectors and relays. Frank Morrell did the engineering design to process the logic. And so we got the alpha test of an automation machine they called the Tunny. The start positions were plugged in by hand and it could still take weeks to decipher messages.

Max Newman, Turing’s former advisor and mentor, got tapped to work on the project and Turing was able to take the work of Polish code breakers and others and add sequential conditional probability to guess at the settings of the 12 wheels of an Enigma machine and thus get to the point they could decipher messages coming out of the German navy on paper. No written records indicate that Turing was involved much in the project beyond that.

Max Newman developed the specs, heavily influenced by Turing’s previous work. They got to work on an electro-mechanical device we now call the Heath Robinson. They needed to be able to store data. They used paper tape - which could process a thousand characters per second using photocell readers - but there were two and they had to run concurrently. Tape would rip and two tapes running concurrently meant a lot might rip.

Charles Wynn-Williams was a brilliant physicist who worked with electric waves since the late 1920s at Trinity College, Cambridge and was recruited from a project helping to develop RADAR because he’d specifically worked on electronic counters at Cambridge. That work went into the counting unit, counting how many times a function returned a true result.

As we saw with Bell Labs, the telephone engineers were looking for ways to leverage switching electronics to automate processes for the telephone exchange. Turing recommended they bring in telephone engineer Tommy Flowers to design the Combining unit, which used vacuum tubes to implement Boolean logic - much as the paper Shannon wrote in 1936 that he gave Turing over tea at Bell labs earlier 1943. It’s likely Turing would have also heard of the calculator George Stibitz of Bell Labs built out of relay switches all the way back in 1937. Slow but more reliable than the vacuum tubes of the era. And it’s likely he influenced those he came to help by collaborating on encrypted voice traffic and likely other projects as much if not more. Inspiration is often best found at the intersectionality between ideas and cultures.

Flowers looked to use vacuum tubes where the wheel patterns were produced. This gave one less set of paper tapes and infinitely more reliability. And a faster result. The programs were stored but they were programmable. Input was made using the shift registers from the paper tape and thyratron rings that simulated the bitstream for the wheels. There was a master control unit that handled the timing between the clock, signals, readouts, and printing. It didn’t predate the Von Neumann architecture. But it didn’t not.

The switch panel had a group of switches used to define the algorithm being used with a plug-board defining conditions. The combination provided billions of combinations for logic processing. Vacuum tube valves were still unstable but they rarely blew when on, it was the switching process. So if they could have the logic gates flow through a known set of wheel settings the new computer would be more stable. Just one thing - they needed 1,500 valves! This thing would be huge!

And so the Colossus Mark 1 was approved by W.G. Radley in 1943. It took 50 people 11 months to build and was able to compute wheel settings for ciphered message tapes. Computers automating productivity at its finest. The switches and plugs could be repositioned and so not only was Colossus able get messages decrypted in hours but could be reprogrammed to do other tasks.

Others joined and they got the character reading up to almost 10,000 characters a second. They improved on the design yet again by adding shift registers and got over four times the speeds. It could now process 25,000 characters per second.

One of the best uses was to confirm that Hitler got tricked into thinking the attack at Normandy at D-Day would happen elsewhere. And so the invasion of Normandy was safe to proceed. But the ability to reprogram made it a mostly universal computing machine - proving the Turing machine concept and fulfilling the dreams of Charles Babbage a hundred years earlier.

And so the war ended in 1945. After the war, The Colossus machines were destroyed - except the two sent to British GHCQ where they ran until 1960. So the simple story of Colossus is that it was a series of computers built in England from 1943 to 1945, at the heart of World War II. The purpose: cryptanalysis - or code breaking.

Turing went on to work on the Automatic Computing Engine at the National Physical Laboratory after the war and wrote a paper on the ACE - but while they were off to a quick start in computing in England having the humans who knew the things, they were slow to document given that their wartime work was classified.

ENIAC came along in 1946 as did the development of Cybernetics by Norbert Wiener. That same year Max Newman wrote to John Von Neumann (Wiener’s friend) about building a computer in England. He founded the Royal Society Computing Machine Laboratory at Victory University of Manchester, got Turing out to help and built the Manchester Baby, along with Frederic Williams and Thomas Kilburn. In 1946 Newman would also decline becoming Sir Newman when he rejected becoming an OBE, or Officer of the Order of the British Empire, over the treatment of his protege Turing not being offered the same. That’s leadership. They’d go on to collaborate on the Manchester Mark I and Ferranti Mark I. Turing would work on furthering computing until his death in 1954, from taking cyanide after going through years of forced estrogen treatments for being a homosexual. He has since been pardoned post

Following the war, Flowers tried to get a loan to start a computer company - but the very idea was ludicrous and he was denied. He retired from the Post Office Research Station after spearheading the move of the phone exchange to an electric, or what we might think of as a computerized exchange.

Over the next decade, the work from Claude Shannon and other mathematicians would perfect the implementation of Boolean logic in computers. Von Neumann only ever mentioned Shannon and Turing in his seminal 1958 paper called The Computer And The Brain. While classified by the British government the work on Colossus was likely known to Von Neumann, who will get his own episode soon - but suffice it to say was a physicist turned computer scientist and worked on ENIAC to help study and develop atom bombs - and who codified the von Neumann architecture.

We did a whole episode on Turing and another on Shannon, and we have mentioned the 1945 article As We May Think where Vannevar Bush predicted and inspired the next couple generations of computer scientists following the advancements in computing around the world during the war. He too would have likely known of the work on Colossus at Bletchley Park. Maybe not the specifics but he certainly knew of ENIAC - which unlike Colossus was run through a serious public relations machine.

There are a lot of heroes to this story. The brave men and women who worked tirelessly to break, decipher, and analyze the cryptography. The engineers who pulled it off. The mathematicians who sparked the idea. The arrival of the computer was almost deterministic. We had work on the Atanasoff-Berry Computer at Iowa State, work at Bell Labs, Norbert Wiener’s work on anti-aircraft guns at MIT during the war, Konrad Zuse’s Z3, Colossus, and other mechanical and electromechanical devices leading up to it.

But deterministic doesn’t mean lacking inspiration. And what is the source of inspiration and when mixed with perspiration - innovation? There were brilliant minds in mathematics, like Turing. Brilliant physicists like Wynn-Williams. Great engineers like Flowers. That intersection between disciplines is the wellspring of many an innovation. Equally as important, then there’s a leader who can take the ideas, find people who align with a mission, and help clear roadblocks. People like Newman. When they have domain expertise and knowledge - and are able to recruit and keep their teams inspired, they can change the world. And then there are people with purse strings who see the brilliance and can see a few moves ahead on the chessboard - like Churchill. They make things happen.

And finally, there are the legions who carried on the work in theoretical, practical, and in the pure sciences. People who continue the collaboration between disciplines, iterate, and bring products to ever growing markets. People who continue to fund those innovations. It can be argued that our intrepid heroes in this story helped win a war - but that the generations who followed, by connecting humanity and bringing productivity gains to help free our minds to solve bigger and bigger problems will hopefully some day end war.

Thank you for tuning in to this episode of the History of Computing Podcast. We hope to cover your contributions. Drop us a line and let us know how we can. And thank you so much for listening. We are so, so lucky to have you.


(OldComputerPods) ©Sean Haas, 2020