Last week, I wrote about visiting the new history of computing exhibit at the Computer History Museum in Mountain View, California and mentioned seeing an exhibit about the ancient Greek Antikythera mechanism. (See “You say you want a revolution? The Computer History Museum reopens after $19M of polishing. Thank you Bill Gates.” ) The Antikythera mechanism (named for where the object was found in a shipwreck) is a sophisticated astronomical computer that predicts the location of significant celestial objects using a complex calculating mechanism consisting of more than 40 metal gears. Some immensely skilled craftsperson built this mechanism about 1400 years before the earliest known European geared clockwork mechanisms. It is an archaeological wonder of metallurgy, mechanical engineering, and mathematics. The mechanism tackles complex calculations that would not again be attempted by calculator designers for perhaps 15 to 19 centuries.
This is a perfect example of the cyclical historic nature of system-design problems that I mentioned in that previous blog entry. Many such system-design problems must be tackled over and over again as our engineering prowess advances and we develop new instrumentalities for attacking apps-driven problems. There’s a clear benefit to studying these older design attempts (successful or not) because engineers, scientists, and mathematicians working with cruder implementation technologies must be craftier to meet their objectives. Today, we’d throw a 99-cent microcontroller consisting of several million transistors at this problem. Back then, a skilled metal craftsperson spent an estimated year assembling the device out of a few dozen gears and some other hand-made brass parts.
As recently as 40 years ago, when developing the first Apollo guidance computer, we threw single-gate Fairchild RTL (that’s the original meaning of RTL—Resistor-Transistor Logic—not today’s “Register Transfer Level”) ICs (4100 of them!) and rope memory (that’s read-only, magnetic-core memory—rarely seen in a commercial product except for HP’s 9100A desktop calculator, circa 1968) at the problem. The Fairchild ICs, all 4100 of them, were connected with Wire-Wrap wiring and then cast into a monolithic block of epoxy for ruggedizing. You’ll have to decide whether single-gate ICs and rope memory potted into an epoxy block are closer to brass gears or to billion-transistor SoCs. Me, I’m voting the gears.
In any case, celestial navigation has been a solved application now for more than 2000 years. Yet the technology marches on and new advances allow us to solve problems in new ways just as GPS has revolutionized terrestrial navigation using speed-of-light technologies and cheap consumer electronics with picoseconds resolution, a system undreamed of 100 years ago.
It should not surprise you that the EDA360 vision is all over this idea. Applications drive system design and they drive SoC design. More than 2000 years ago, celestial navigation drove the development of a surprisingly complex mechanical calculator that would not be equaled in complexity for another 15 centuries or more. In the 1960s, the same sort of application drove the development of then-new IC technology like no other application then in existence.
Do not underestimate the power of apps-driven design. Given the technology, you can rest assured that someone will come up with a surprisingly ingenious solution.
If you’d like to know more about the Greek Antikythera mechanism, the National Geographic Channel has devoted a segment of “Naked Science” to people working on unraveling the mechanism’s complexity and mystery. I wish I could replicate the videos here on this blog, they’re extremely interesting, but wordpress.com won’t allow it. So here’s the link to four short segments about the mechanism:
Be sure to watch all four. The full story will air on January 20, that’s three days from now. Have fun!