Love this August, 1935 article from Science and Mechanics magazine about the hulking, three-ton ballistic computers -- reminds me a lot of the way that Asimov and Heinlein both wrote about "big brains" in their fiction over the next couple decades:

Link

The "fire, control" machines, now used to plot the flight of shells from modern guns in moving ships, against moving targets, deal with practical conditions like this; and the machine pictured could answer a question of this nature, as well as a good many others less specialized. For instance, three or more heavenly bodies (like Earth, Sun, and Moon) are moving in their orbits at different rates of speed and varying distances, attracting each other. What will be the combined result of their forces, in changing the positions of each, in a given period? It is an enormously difficult proposition for the best mathematician in the world. With this machine, its ten "integrators" would be adjusted (by setting dials) to represent the varying factors of the problem, and then started turning. The friction discs and gears of the machine would operate on each other, each of them with an effect proportioned to the energy and speed it represented; and, on the final chart at the "answer table" of the machine (see illustration) a curve would be drawn by a metal pen, representing the formula desired (not necessarily a physical picture of the motion of one of the heavenly bodies, but a mathematical picture of it).

I remember back in the mid-’80s taking a tour of an active-duty battleship while in the Navy. It used huge electro-mechanical computers for the fire control of its big guns. They took up a substantial part of the ship itself, and were definitely “old-school” even at that time. But- according to the officer showing us around- modern computers weren’t any more accurate, so they had no reason to replace them.

They worked really well. Until you got your tie caught in it.

One interesting point overlooked: This is an analogue computer, today these problems are solved by digital computers.

Yes, Sae Miller is correct, this is a Differential analyser, which was analog. This was a pretty advanced one, but the first had been around a few years prior. Meccano.us is a neat site to look at for more info. Meccano look cooler than Legos definitely, can’t build your own Difference Engine with Legos. Perhaps you can, I’m just not that clever.

See, the Heinleins didn’t have a big enough garage nor enough electricity available for one of these, that’s why

“writing the description of how they got to that station required the famous three days of paper-and-pencil calculations (see Expanded Universe, pp. 519-520) for one line of writing. John Campbell would have been pleased to learn of this, but Heinlein was now writing for an editor who did not appreciate such effort.”

http://members.iglou.com/jtmajor/SpaCadet.htm

I recall somewhere Heinlein describes the same kind of effort, days with a pencil and paper, doing the calculations for “Destination Moon.”

Clearly – the analog computer eventually evolved into the Fusbol Table.

Solving the earth-moon-projectile gravity thing with a mechanical device is really interesting. I had heard of a ballistics type computer. But I always heard it described as artillery trajectories, and it seemed like a table of trig functions would have worked for that. This makes much more sense.

I had to implement the Runge-Cutta Feldberg (I think) algorithm in Fortran for a class one time to generate a set of plot points for a projectile leaving the earth and passing the moon. I don’t recall what order differential equation it was, but having to compute each point by hand would definitely be a week long endeavor.

You can larf at them now, but the main problems with them is setup of a problem, aka “programming”. Once done, however, many analog machines “compute” instantaneously. Besides the obvious bulk, and setup hassles, accuracy is limited to 3? or so decimal points.

THe scalability of precision is what made digit-based calculations attractive (keep in mind that automatic, electronic, digital, computers were much, much larger than mechanical calculators for over a decade).

Improvements in analog precision gets very expensive very fast. 5 digits is more or less resolving 10 microvolts out of one volt; 6 digits is one microvolt; most amplifier noise is in the tens of microvolts range. The cost scale is geometric.

To double digital precision you simply build another row (or column, …) of the same stuff as before. Cost scale is linear.

Analog modeling is really attractive in some ways; when done right, it’s equivelant to parallelism still unthinkable in digital terms. And often the real-time nature is more important than accuracy, especially in closed-loop problems.

My old head feels heavy. Somewhat sorry for the short comment, but I’m playing with my new Asus Eee, which has no vacuum tubes, but is cool as all hell, hard to type on, but it beats flipping switches for each card. lemme see, colon right parend….. :)

This thing looks a bit like a horizontal Jacquard loom. We got punch cards from Jacquard looms – that’s how their designs were “programmed” in. There’s an interesting PBS show, pretty old at this point, of how cotton farming in Egypt gave rise to the modern computer.

Farming -> clothing -> cotton -> weaving -> trade -> trade routes -> fiefdoms – protection of trade routes -> castle/forts -> -> seiges; longbow -> crossbow -> catapult; gunpowder -> blunderbus -> cannon -> ballistics -> complex math to solve parabola of projectile -> this analog blunderbus above -> ENIAC; cotton -> cloth -> hand looms -> mechanical looms -> Jacquard looms -> punch cards to program Jacquard loom designs -> ENIAC.

And then there was COBOL which begat…

If I have any bits (no pun) wrong, feel free to correct me.

Feinman goes into some detail about

his work on mechanical balistic computers

in “Surely you’re joking, Mr. Feinman”

#8 — I once worked with an engineer that had worked in an aircraft factory in WWII. He described the process of doing a stress calculation as:

1. Engineer sets up the problem with the appropriate numbers.

2. Papers are passed to another section that converts all the numbers to logarithms and corresponding necessary addition/subtraction problems.

3. Results checked for accuracy, then are attached to the original papers and passed to a third section.

4. This section is full of women (it was WWII) with mechanical adding machines. Here they did all the adding and subtracting of the logs.

5. Results were checked for accuracy, attached to the paperwork and returned to the previous group.

6. Results were translated back to real numbers from the logs and checked for accuracy.

7. Results were then passed back to the engineer.

This normally took about a week to do unless you had full priority which you “might” get your results in 3 days.

Remember, this is for ONE loop around the design process; if it wasn’t strong enough, you got to do it again. Also, there was no “cut-and-paste” for entering the numbers. You had to key in each number every time you wanted to perform a calculation. If the adding machine papertape was lost, you got to do the whole calculation all over again. Imagine trying to optimize something!

Obviously, this process wasn’t used for all their calculations. They had their sliderules, hand calculations, and engineering handbooks for quick-and-dirty stuff, but this was the general process for the large critical calculations.

-eef

You can actually see one of these machines at work in the 1956 film â€œEarth vs. the flying saucersâ€ (at â€“more or less- 56 minutes from the start). The data output device shown on the film is graphical, but seems, however, very advanced (even by the present standards):).

In the film you can also see the panels in the walls to store the little wheels used to â€œprogramâ€ the thing.

A modern and extremely beautiful recreation of one of these computers was done by Tatjana van Vark. ( http://www.tatjavanvark.nl/harmonium/ )

Seems SteamPunk, but is a completely different level.

The photo appears to be a “ballistic” rather than a “fire control” computer. A ballistic computer is used to come up with the tables and profile the cams that would later be used to fire artillery. You have to figure out what the constants are for x shell with y charges at z temperature before you can write the firing tables and machine the fire control computer.