Thursday, April 29, 2010

Steampunk heaven

This one's another not-particularly-on-the-web item, though you can, of course, learn more about it on the web.

Over a hundred and fifty years ago, Charles Babbage designed his Difference Engine (actually, he designed it twice, the second design requiring many fewer moving parts). Less than a decade ago, it was finally built. One can be seen at the Computing History Museum in Mountain View, CA, at least until its owner finishes reinforcing his living room floor to accommodate it (hey, you can never have too many five-ton Victorian computing machines in the house).

I remember reading about the difference engine many years ago. The wisdom at the time was that it could not have been built to acceptable tolerances with the technology of the day, or if it could have been, it would have required too much force to operate.

Both of these notions have been disproved, and in fact two working examples have been built, using modern numerically-controlled milling machines, but with carefully-introduced errors to mimic the imprecision of the time (not as much as you might think -- the Victorians knew a bit about machining). With some training and practice, and with the aid of a 4:1 reduction not in the original plans, a person of average physique can literally crank the thing up and watch it produce values of a seventh-degree polynomial to thirty-one digits.

It's something to see. Forget about retro PC keyboards, this is the real deal.

Of course, the real fun would be in cranking up a Babbage Analytical Engine, but alas, the design for that one is not complete enough to implement. Not that that stopped Ada Lovelace from hacking it anyway.


earl said...

During WWII the navy used mechanical analog computers to control guns. Multiplication and division were handled with differentials (in the mechanical, not the mathematical sense), and more complex operations were done with specially ground cams, or so they told me.

David Hull said...

That's my understanding, as well. Lots of interesting but largely forgotten math in that. Functional square roots, for example.

The interesting thing about the difference engine, to modern eyes at least, was that it was digital, very explicitly so. For example, there is a sharp-edged bar that slots in at certain key points to make sure that the number wheels are exactly in one of their ten allowed positions.

Babbage's analytical engine, unfortunately not quite fully-baked enough to build, went even further, anticipating a fair bit of Turing, Von Neumann and company's work.

The intriguing question here is what would have happened had Babbage's digital designs succeeded? Instead, there followed a long period of analog dominance. Digital processing wasn't dead -- I'm pretty sure punch cards have a pretty continuous history from the Jacquard loom up to the 60s and 70s, for examle -- but it wasn't the dominant force it is today.