Slashdot Mirror
Building Babbage's Analytical Engine
An anonymous reader writes "Anybody think 3-D printing technology will have enough moxie to pull off a construction of Babbage's analytical engine by 2021, the 150th anniversary of Babbage's passing? The Guardian reports, 'Plan 28 – named after one set of Babbage's plans – has assembled the leading technical experts on his designs and just started fundraising. The first stage of the project involves studying the thousands of pages of handwritten notes that Babbage left behind, to determine what exactly needs to be built. Once the study is complete, we'll be building a 3D physical computer simulation of the analytical engine to verify that his design is workable. Reaching that stage is likely to cost about £250,000. Only once the feasibility of building the machine has been established will the much larger fundraising effort needed for the actual construction to begin. But what we hope to do is create a working monument to the man who conceived the computer, and to inspire today's scientists and engineers to dream a century into their future.'"
Another rebuild that has progressed way further: http://www.meccano.us/analytical_engine/index.html I suspect this will end up a lot cheaper too....
That project is starting to sound like a boondoggle. Lots of PR and fundraising, no hardware. They have a contribution system, a mailing list, a Twitter feed, and press coverage. They've been blithering about this for two years now. But they haven't built so much as one single demo part.
We know what the Analytical Engine was supposed to do computationally. There's a simulator. It's a rather straightforward machine. It's roughly comparable to a programmable calculator of the 1970s. There are 1000 memory locations, each of which stores a 50-digit decimal number. These are separate from the program and data, which are on chains of punched cards. It can add, subtract, multiply, divide, shift, and compare, which is all you need.
Parts of the Analytical Engine have been built, and there's a working Difference Engine. So the components are understood.
There's no good reason for the 50-digit precision, and 1000 memory locations is too much for the compute power available (about 1 IPS). Like programmable calculators, 10 digits and 100 memory locations would have been enough for most problems. Babbage's own trial model of the "mill" (the ALU) has only 25 digits. Building a memory of 50,000 wheels about 3 inches in diameter means building something the size of a locomotive, most of which will just sit there. Trimming it down to 25 digits and 100 locations would make for a large desk-sized machine.
A question I once asked of the project was "how many part numbers"? That is, how many different parts are required? They didn't know. I suspect not that many. The existing model of the mill doesn't have a high part number count. The "store" (the memory unit) is inherently repetitive. Most of the parts can be die-cast and finish-machined, which is the most economical way to produce good metal parts in medium quantity. Many of the lever-type parts are cut from flat sheets of brass. Those you make with a CNC mill or a water jet cutter. 3D printing isn't really appropriate as a way to make brass parts, and making a plastic copy of the Analytical Engine would be rather tacky.