How a 1950s ballistic computer worked

Here's a series of 1950s-vintage videos illustrating the workings of mechanical Navy ballistics computers -- enormous arrays of switches and gears that are used to quickly and reliably do complex mathematics in response to changing wind, targets and other factors. It's a really good, easy-to-follow guide to the underlying logic in a computer, and it's excitingly mechanical.

u.s navy vintage fire control computers (part 1) (Thanks, Harold D and everyone else who suggested this!)


  1. Wow. WOW. Lookin’ at the redesign (as of 8:30am EST) and it’s wonderful! (Sorry for being OT, but I can’t find the thread for redesign comments ;-) I was not a big fan of the previous two overhauls; this one makes up for both of them. The layout is clean; text scans very well; the sidebar now has something to say other than just, “Look at me”…heck, the pages even load in a reasonable amount of time — did you guys finally discover iFrames? Yay!

    I know it’s not done yet — like, this comments input box should open to column width; the surgeons still need to implant a search box; the “<!– output divh –>” section needs fixing; CSS needs small tweaking; etc. — but this new page is several giant steps forwarder for the site. Well done, all!

  2. I absolutely love that the 16mm projection mechanism is audible throughout. Sweet, sweet music!

  3. this is so great, I’ll use it in my courses (I’m teaching physics to futur machining tech,)

    Mechanical integrations of coupled differential equations… miam!

    If you’re into those things, you probably already know Van Vark…

    She has updated her “ball and egg” personnal picture… 8-)

  4. I read a fantastic (but somewhat technical book from MIT Press on this a couple of years ago:

    A compendious discussion of how the introduction of long-range artillery at sea was a necessity in the age of steam driven ships, and how this introduced problems of leading – compensating for several simultanously changing vectors – that defeated artillerymen in the period before the 1st World War.

    This discussion progresses through increasingly complex machines, with higher and higher degrees of mechanization, leading to the introduction of powered gyroscopes, integration mechanisms, aiming and tracking devices, mechanical storage of tables, etc.

    Starting with big guns, this then moves to the problem of tracking a moving aircraft from a moving, rocking platform like a ship: the maximum problem arises when Japanese pilots fly aircraft into the vessel.

    A parallel narrative discusses the rise of information theory in telephone systems, the introduction of multiplexing, and how the theory developed there came to describe the limit cases for mechanical computers. And we all know what happened then.

    A bit like Andrew Hodges’ bio of Turing, it’s really a great way to understand and put in perspective the mechanistic underpinnings and assumptions of the early cyberneticians, and showcases how sophisticated their ideas about information could be within a mechanical system.

    Not for non-geeks, but great.

  5. my grandfather developed a mechanical aiming device for dive bomber airplanes back in the 1940s that worked like a threedimensional curved slide rule.

    pretty neat stuff, genious from a maker’s perspective. flying these airplanes in low-altitude missions gave him more than enough bad dreams in his sunset years.

    i think i will never understand how people can sell their ingenuity to killing people on a massive scale. in retrospect, neither did my grampa.


  6. Impressive!

    My daughter, a freshman in high school, is studying geometry this year. She’s familiar with similar triangles and the proportionality of their sides. She’ll appreciate this!

  7. This is great. I think I speak for all mechanical engineers when I say this sort of thing makes me intensely happy. I’m going to look up the other parts of the series.

  8. It strikes me that analog computers would be useful for teaching math. There’s something about seeing the differences (even if scaled.) The problem with calculators and computers is that they’re black boxes.

  9. Analog computing. This dates to the time when one digital logic element was the size of a salt shaker. Now you need an electron microscope to see same.
    I’d love to see the details of the Norton bomb sight. Same range of problems.

  10. Here is a book on the subject. He’s a terrible writer, needing a good editor — he gets so carried away that pretty soon he’s using terms several chapters before they are defined. The title is misleading. The book could have been a lot better, but he does get into some good historical info on how these things developed. I haven’t seen the other recommended books, don’t know how they compare.

  11. Saw one of these things the other day at a NATO museum here in Denmark. Gigantic thing coupled to a couple of WW2 cannons the Danish confiscated from the Germans and was meant to stop the Russians from sailing from the North Sea into the Atlandic.

    It had a million dials and levers and could supposedly hit within a 10m radius of its target. Really really cool.

  12. The great thing about analogue computers is that they can be perfectly accurate to the limit of physics. In the given example, the output is exactly the product of the height of one input by the offset of the other from neutral: as expressed in photon widths or whatever you please. Your problem then is precision: how exactly can you set the inputs to correspond to the values you care about; how accurately are you reading the results?
    So your error becomes the noise or inaccuracy in your transduction of the real-world problem into the realm of the machine and then back out. Until recently the transductions of the analogue computers were easier to accurately set up but the digital ones gave more precision – that is to say more digits in the answers. In the gunnery example the analogue input of the vectors of the muzzle position are input with rods directly attached to the gun and your error is just the flex in the rods – practically nothing. But your readout is something like “somewhere between the “4” mark and the “5” mark. A digital solution varies some current as the position of the gun changes and your error sources are many and include interference from all of your local electrical phenomenon and the accuracy of your amp meter. But your readout is a snazzy 4.33212974372947389
    Digital eventually won out as its machinery doesn’t wear out as fast and is now cheaper plus the fact that it can calculate faster then the phenomenon it’s measuring can change so you can do forecasts. Analogue is limited to the speed of the real world.
    One of the things I love about Steam Punk is that it posits a world where we stuck with analogue computers!
    The British Navy had, at one point, a very sophisticated hydraulic computer used in its submarines. One of the inputs and its only power source was the outside pressure of the water. Ontario Hydro (local power grid) used to have a table top model of the high tension corridors in the province only with little voltages. Anything you did to it would give the same results as if you were doing it to the big grid: even unexpected results not covered by your equations. You could then ponder the data and refine your equations without the mess of exploding sub-stations.
    The neatest one I ever ran across was the huge, “Emperor Ming’s Space Organ” controller for the Sarnia to Toronto pipeline system here in Ontario. A beautiful brass and steel colossus that directly pulled in pressures from the pipelines, directly measured the densities of the gases therein, whirled the data through some manifolds and spat out pressures that spun valves to give the desired results out in the field.

    1. In the given example, the output is exactly the product of the height of one input by the offset of the other from neutral: as expressed in photon widths or whatever you please.

      Well, sort of. If you presume that the construction of the machine is as flawless as the abstract math it is modeling, then yes, the output is exact.

      In practice you have to account for error resulting from imperfections in construction and materials. There is backlash in the gears, slop in the pin to slot fitting, waviness in the slots, shifting dimensions and linearities resulting from shifting stress on the parts, changing temperatures, etc.

      Cool stuff, no doubt, but not exact.

  13. Nadreck #15, the biggest error will be the operators estimation of the input variables. Angle is relatively easy but range? The initial range must be accurate or all subsequent calculations will carry forward the same error. In addition, current course and speed are constantly changing. The machine is incredible in that it is analogue and constantly calculates the output (as long as the inputs can follow the changes). Modern computers appear to do the same only because the time steps are so small as to be indistinguishable, but they are still there.

    One question not answered in the video is how accurate does the whole machine need to be to achieve an acceptable hit rate? Including the necessary measurements and operator inputs. I especially liked the bit where three operators are twirling the dials. Anybody know of any videos showing it being used in anger?

  14. Brings to mind the brilliant and elegant Curta mechanical calculator. It’s mesmerizing.

  15. Nadreck and GraemeM–And of course the person estimating the inputs is often NOT the person inputting the data into the computer. Certainly when using the Torpedo Data Computer, the submarine equivalent, the Captain would call out the target data, it would be echoed by the operator to make sure he was inputing the same data. So, submarine course and speed which were automaticly updated in analog, the target data (distance, angle on the bow etc) WAS not analog because it had to be communicated and input.

Comments are closed.