Escher.gif (426 bytes) FTL Design
Babbage's Difference Engine (1822 - 1991)

The Author Declines to Turn the Crank
by Carrie Dougherty

Editor's note: Carrie is a civil engineer who lives in Oakland. Although, how civil can an engineer be who once wrote, "I am a civil engineer, if it doesn't work, I hit it with a hammer." She also wrote, "Who would have thought I'd be the proud owner of a motorcycle that thinks it's a duck." Referring, of course, to the antique Royal Enfield J2 she bought and drove around England before she shipped it home to Oakland, California. Which has not stopped Carrie from going back to England several times and continuing her motorcycle tours of that country on a rented motorcycle, if necessary.

You see, Carrie does not drive automobiles. But she is a member of LASFAPA and she very entertainingly writes of her various adventures. Her very inquisitive adventures as she does more than motor through the countryside. And continue to look for a job in England.

This article is excerpted from her August 2002 zine in LASFAPA #311.

Carrie writes of her visit to the world's first computer, the 1832 Difference Engine invented by Charles Babbage, begun in 1822, suspended in 1842, and finally completed at the London Science Museum for Babbage's 200th birthday in 1991.

(Introductory notes by Marty Cantor, from No Award #12)

If you're interested in Carrie's further adventures (or know of a job available in England), visit her website.

The Author Declines to Turn the Crank

London contains hundreds of fascinating museums, galleries, churches and stately homes, the vast majority of which I've never seen, and a few of which I've only seen once, briefly. So, naturally, I did what I always do in London and headed straight off to be at one with the machines in the Science Museum. The name, by the way, is somewhat of a misnomer, as the museum is principally one of historical technology and engineering. I headed straight upstairs to the Difference Engine, where chatting with the two technicians elicited the information that they would be testing the Engine in the afternoon, after their lunch break. I promised to come back later.

So I spent an hour and a half or so wandering around, seeing maybe 1/100th of what's interesting in the museum, including the following:

• The Great Hall of Engines (formally called East Hall), which has magnificent specimens of 18th century Newcomen engines as well as a beautiful gigantic Corliss engine which apparently runs now and then (although I don't believe I've ever seen it operating).

One thing I learned this time was that the cycles of the old engines were very slow—maybe one or two revolutions per minute.

• The new Making of the Modern World gallery, which contains "iconic" artifacts from the history of science and technology, including a V2 rocket (I'd never realized how huge they were), the original Rocket locomotive, a Cooke and Wheatstone telegraph station that looks like an organ console, and a phrenology head (apparently on loan from the Phrenology Head Museum in Cornwall - -!)

• The King George III Gallery of 17th and 18th century instruments, mostly teaching tools and "parlor experiments" used for recreation during that era. It has the original orrery built for the 4th Earl of Orrery in the early 18th century (though there were actually earlier Chinese and Roman ones, apparently). The earliest orreries showed just the earth, moon, and sun –you can tell the age of more elaborate orreries by how many planets they show, unless they've been altered to show additional planets. And where else can you find anamorphic pictures (hah, spellchecker doesn't like that one, nor did it like "orrery"), scioptic balls, and cometaria?

• A model and the original mirror of the Great Rosse Telescope of Birr Castle, Ireland. I'd heard of it in conjunction with my visit to William and Catherine Herschel's house in Bath last year, but hadn't realized that this 72" telescope, the largest ever made when it was constructed in the 1840s, has been fully restored and is once again operational.

• The Phillips economic computer, a set of clear plastic sluices and adjustable valves (labeled "taxes," "savings," "consumer spending," etc.) which mimics the actions of a national economy when you pour colored water through it.

Charles Babbage's brain
- full description

When I came back the technicians had opened the glass case that the Engine is kept in, and I settled down to observe the test. I would have been perfectly content to watch this incredible collection of wood, brass, and steel go through its paces, but pretty soon I was behind the velvet rope with the technicians, looking over their shoulders, inspecting the results, and making (no doubt helpful) suggestions. Richard soon left, and Reg graciously spent another couple of hours answering my questions and pointing out interesting features of the mechanism until we finally said goodbye just before the museum closed.

The printer now works, and almost all of the tweaking is done. That afternoon Reg and Richard were testing the way the printer makes impressions into a heavy plaster of Paris plate, which would have been used as a positive mold to create lead type plate to print mathematical tables. The numbers were now more even than they had been, but the impressions appeared to vary in depth. Richard said that this was a flaw not in the machine but in the plate-the plaster has to be both flush to the metal frame of the plate and completely flat. They had experimented with many mixtures of plaster and degrees of hardness, having received no direction from Babbage and with no idea how he would have done it. They were still working out the appropriate medium-when I suggested some kind of metal foil they pointed out the risk of disaster if something metallic attached itself to the number faces of the printer; plaster was easy to clean off. Many people, they said, had told them what not to do but no one had yet come up with a good answer.

The printer has two typefaces, large and small, connected with flexible metal strips that allow force to be conveyed to either typeface in an arrangement that Reg called "pure Babbage."

When they built the printer they extended the main shaft that lifts the cams that drive the columns, but also added a feature that Babbage hadn't thought to include, the ability to disengage the printer from the shaft and an additional crank on the printer that runs the printer only. The number columns themselves can't be disengaged from the printer mechanism, but the printer can be run without the engine, and will just print zeros.

The printer can print results either in two columns or directly across the page, though Reg can't figure out why anyone would want to do that because it takes so much effort to push the frame back and forth inside the printer. Reg showed me where he installed an oil-filled piston to damp the momentum of the sliding frame to keep it from being propelled right off the edge of the printer (i.e. manual line break). He also showed me how a cylindrical weight drops out of a bracket when the printer reaches the end of the frame, releasing the tension in a wire that connects to the crank (i.e. manual page break). This weight was falling off too soon the afternoon I was there, and needed some kind of adjustment to stay on until the actual end of page.

DSCN2409e.jpg (63364 bytes)As I watched, Richard and Reg spelled each other on the crank. It apparently takes a great deal of technique to turn the crank, and in fact I saw Richard jam it at one point by inadvertently slowing down before stopping. It's necessary to turn the crank at the exact right speed, neither too slowly nor too quickly, and not to accelerate or decelerate. This is challenging because the load on the crank varies during the engine's four beat cycle (the load increases when the locking blades lift out of the number wheels). Given that, it seemed to me that the machine might run better if operated by a machine—but a machine wouldn't have the ability to continuously monitor the workings of the engine by feel.

The engine can be driven by one person, though it's tiring; Reg altered Babbage's design by installing a 4:1 differential gear at the crank which means the operator faces in the opposite direction from what Babbage had intended. It may be possible for someone to operate the engine with four times the force, but it would be more difficult to turn the crank at a steady rate during variation on the load. It was amusing to listen to the two of them discussing the machine as if it were some kind of ordinary gadget—"give it a few more turns, will you? Try taking it up to 35"—and to watch them tune, tighten, and adjust it, using flathead screwdrivers and a small crowbar, for all the world as if it were an old British motorcycle.

The differential gear is one of the many minor modifications to the original design that Reg made in the course of building the engine and the printer. He showed me a few others, mostly on the printer.

One thing Babbage seems to have neglected was leaving time for operations to complete themselves. There originally wasn't any time in the cycle, for example, for the thin metal strips that maintain the locations of the number wheels to move in and out of position between the teeth of the wheels. Obviously, they need to move out, along a vertically inclined plane, to allow the wheels to turn, then need to slide into place once the calculation has been made (causing a characteristic crashing/clunking sound). In order to give the wheels time to turn, Reg had to cut some sharp notches into the wheels that turn the shafts that lift and lower the columns and locking strips. Babbage also didn't include any counterbalances for the weight of the locking strips, though he had included springs to balance the weight of the columns; Reg had to figure out a way to add these himself.

DSCN2408e.jpg (99287 bytes)One interesting and effective design feature of the Engine is that when any given piece isn't moving it's locked into place, which both prevents damage to the parts and prevents the machine from making errors—"either it gives the correct answer or it jams." The only problem with this is that the carry hammers sometimes slam into locked number wheels and break off; Reg handed me a shoebox containing half a dozen or so of these. These are the only parts of the machine that are actually subject to breakage. It apparently wasn't always the case that the machine always gave the correct answer, though; initially they had had a problem with the carry mechanism, and it was Reg's task to try to determine which of the 200-odd warning levers for the carry hammers wasn't turning.

Another place Babbage didn't leave any time for things to happen was in the printer. Aside from making printing plates, the Engine can produce results on strips of paper. When the number stamps on the wheels fall into place, a roller spreads ink on the type, then a paper roll is pressed up against it. Apparently, though, there wasn't enough time for the ink roller to get out of the way before the paper roller comes up, which must have been a great mess. To solve this problem Reg created a counterweight arm with a groove in it that controls the movement of the ink roller, snapping it smartly out of the way at the end of its arc. He referred to this as "a bit of Heath Robinson," Heath Robinson apparently being English for Rube Goldberg.

Reg pointed out that they had decided early on to make whatever modifications were required to allow the machine to function. It was better to have a modified, working engine than an unmodified unworking one, and the modifications are nothing more than what Babbage himself would have had to do in the course of constructing the machine (it's actually remarkable, Reg said, how few modifications he'd had to make, given that Babbage had basically sketched this machine on the back of an envelope). The modifications were designed and installed in such a way that they could be easily removed if for some reason it was necessary to have an "original" machine—the basic design follows Babbage's drawings as much as possible.

Another thing that hadn't apparently crossed Babbage's mind was how the printer mechanism would be supported. Reg said that in a way the printer was harder to build than the engine itself, because all of the pieces had to be fitted together, then intersections had to be marked, then the pieces needed to be disassembled and holes drilled before fastening them together. "Then," he says, "you discover to your horror that the whole mechanism is fastened to the Engine by two screws." After a great deal of discussion, Reg designed a support frame for the printer.

DSCN2407e.jpg (64491 bytes)I learned something about the actual operation of the engine. To set it up, for example, is a laborious multi-step process. The master gear is marked with 50 divisions, in increments of 5, as well as marks for full cycle, half cycle, set odd and set even. To initialize the engine, first release the two levers at the bottom of the machine "odds" (columns 1, 3, 5, and 7) and "evens" (columns 2, 4, 6, and 8). Then turn the wheel to 10, and set the number wheels on the odd columns to 0. Then allow the mechanism to lock by sliding the knife-blade strips between the number wheel teeth—this keeps the wheels from rotating randomly during the rest of the process. Allow the mechanism to lock after subsequent steps as well. Then turn the wheel to 35 and set the even columns to 0. Then turn the wheel past 0 to 20, or Set Odd, and initialize the odd columns by setting the number with the required differences. Then turn the wheel to 45, or Set Even, and initialize the even columns. Then turn the wheel again to 0, pull up the odds and evens levers, then unlock the mechanism preparatory to calculating. (Hmm - computers have not changed one bit. - ed.)

The machine can print 36 numbers in a row—technically this means it can calculate to 36 places, but in practice the first few numbers are used as counters, and zeroes are set between the counters and the calculation, so each series of numbers might look something like this:

240000000000000000000000000000057600
241000000000000000000000000000058081
242000000000000000000000000000058564

I don't know what equation they were calculating that afternoon, but it was a seventh order equation, the highest the machine can calculate. For lower order equations, the columns closest to the crank are not required and the number wheels are set to 0 and don't turn. I wondered how long the machine would have been likely to last if it had been put to its intended use of creating tables, and had been operated continuously for as long as it would have taken. Reg responded to this question in a few different ways. First, the machine would not have been used continuously to print out tables—it would have made only one set, and then the tables would have been printed from this set. Second, as far as they could tell, the machine was actually improving with age. It no longer needs much oiling, for example. Initially they lubricated the columns from the top; I remember when I first saw the engine there was a puddle of oil in a metal tray underneath it. Now that the mechanism has bedded in, however, it no longer needs to be oiled and in fact works better and faster without much lubrication, though they put graphite on some of the moving surfaces. They also found that the drag on the columns could be reduced by milling out grooves in the columns where they weren't supported by or interacting with other parts, to reduce the friction when they move. Now the machine can do a calculation in six seconds when it first took ten; in fact, it now initially calculates faster than a laptop, which is a result no one expected.

Apparently part of the deal arranged for getting funding for the printer was that a second Difference Engine will be provided for Seattle in 18 months. This machine will be assembled at the Science Museum (they're still debating about whether they'll assemble it in a public gallery—Reg, Richard, and I all think they should) and shipped to America. Richard will probably go with it, though Reg, who has already retired once, doesn't want to. I've e-mailed the curator of the exhibit to get contact information for whoever's going to be getting the machine in Seattle; if I'm still in the US by that time I intend to offer my services as an apprentice Difference Engine mechanic—I'd hate for all of this expert training to go to waste.

Text copyright © 2002 Carrie Dougherty. Photographs copyright © 2002 Bill Burns

The images above are of the 1991 Difference Engine.
Also on display at the Science Museum, and shown below, are
parts and assemblies made in the 19th century by Babbage and by his son.

DSCN2413f.jpg (36823 bytes)

DSCN2416e.jpg (55938 bytes)
DSCN2411e.jpg (78907 bytes)
DSCN2414e.jpg (56031 bytes)

The Computer History Museum in Mountain View, California, has on exhibit a small part of the 1991 Difference Engine, together with the second modern Difference Engine.

No longer on display at the museum, but still of particular interest, is Tim Robinson's working model Difference Engine built entirely from Meccano. American readers not familiar with the British Meccano components may see similarities to Erector Set parts.

And on a fictional note, Eileen Gunn's Difference Dictionary website (designed as a companion to The Difference Engine, a novel by William Gibson and Bruce Sterling) gives much historical background on Babbage and his machines.

Technical notes on the Difference Engine

Babbage Difference Engine #2 - How to Initialize the Machine


Page design copyright © 2004
FTL Design,

Last revised: 2 March, 2015