Gregory's ServerThe first Minuteman model used a transistorized computer called the D-17B, constructed as a cylinder to fit the missile. It was a 24-bit computer, but serial, operating on one bit at a time with a one-bit ALU, so it was slow. 7/N
|
55 comments
 This photo shows a transistorized circuit board for the disk read amplifier. The computer used diode-resistor logic and diode-transistor logic to minimize the transistor count. It used 6282 diodes, 5094 resistors, and 1521 silicon and germanium transistors. 9/N Minuteman II used one of the first integrated circuit computers, making it an important contributor to the IC industry. The D-37 computer (in the engineer's hands) was much smaller and somewhat more powerful. It was still a serial computer with no RAM, but it was able to take over many of the ground control checkout and monitoring tasks. 10/N This photo shows the D-37C computer with advanced 10-layer circuit boards covered with Texas Instruments flat-pack integrated circuits. It used 22 custom integrated circuits; a flip-flop chip cost over $500 in current dollars. 11/N Minuteman III used the upgraded D-37D computer with 14,137 words of storage. It used flexible PCBs, advanced for the time, soldered with low-melting-point indium/tin solder. 12/N A nearby nuclear blast could cause transient errors in the computer. To prevent this, a "transient current detector" (photo) sensed a radiation pulse and the computer was shut down until the pulse went away. The computer would then continue where it left off, extrapolating the trajectory for the time it was shut down. 13/N  The Missile knows where it is because it knows where it was ... 😉 The guidance system also contains a stack of amplifier boards, interfacing the computer to the rest of the missile. The amplifier includes high-current drivers for the various pyrotechnics on the missile such as igniting the rocket stages and retrorockets. 14/N Another block of electronics is the Missile Guidance Set Control, which has amplifiers, power supplies, and servo circuits for the gyroscopes and accelerometers. It has a modular construction with a removable module for each function. 15/N The missile is powered by an unusual battery. This silver/zinc holds the electrolyte in a separate tank so it has almost infinite shelf life. At launch, a gas generator is ignited, forcing the KOH electrolyte into the battery cells, powering up the battery. 16/N Another pyrotechnic component is the squib-activated switch, a switch that is activated by a tiny explosive squib. The missile is connected to ground equipment by an umbilical cable, which is disconnected at launch. The explosive switch severs critical signals just before, presumably to avoid any electrical noise when the umbilical is disconnected. Also note the window in the side of the missile for the alignment light beam. 17/N The guidance system runs 24 hours a day so the missile is ready for immediate launch. On the ground, the system is water cooled; ground support equipment provides a chilled sodium chromate solution through the umbilical connection. This photo shows the coolant path through the computer and other components. 18/N The missile is programmed with launch codes through the Permutation Plug, a plug that holds the 48-bit code. With great ritual, an armed officer plugged it into the socket below. The launch code also depended on values on disk, so each Launch Control Center had a unique code for each missile. 19/N A launch normally requires launch orders from two separate Launch Control Centers. But a single surviving Launch Control Center could launch the missiles, unless vetoed before a timeout. A complicated state machine managed the launch process. 20/N Although the Minuteman guidance system is interesting technologically, one has to keep in mind its purpose was to unleash nuclear devastation On the other hand, Minuteman has been successful as a peacekeeping deterrent (so far). In any case, it is morally ambiguous compared to, say, the Apollo Guidance Computer. There are currently 400 Minuteman missiles active, down from a peak of 1000. 21/N I wrote a blog post that goes into much more detail on the Minuteman guidance system and computer, so check it out: https://www.righto.com/2024/08/minuteman-guidance-computer.html  @kenshirriff Two "super stupid questions" .. actually one maybe a bit less stupid .. 1st question "Of course all that ( very expensive ) tech gets destroyed once the missile is launched ? I suppose when doing tests with no explosive they managed to recover it ? 2nd maybe less stupid have they ever had a "self destruct" thing to avoid such tech to accidentally end up in enemy hands if a missile launched failed to explode and could be recovered ? Maybe I am "too tinfoil hat territory" ? 😅 @gilesgoat Since the missile (and warhead) are crashing down at Mach 23 from hundreds of miles up, I don't think there would be anything left to recover. @kenshirriff Thought so, it breaks my heart to think all that fantastic tech destroyed in an instant .. 😢 @kenshirriff came for the electronics, stayed for the typefaces :D so 60s  @kenshirriff The technology is fascinating, and you've explained it very well, but promoting/sharing weapons of war doesn't feel good to me 😅  @kenshirriff @johnefrancis  @kenshirriff @johnefrancis the 1964 Minuteman incident is written up here: https://www.armytimes.com/news/2017/11/04/details-of-south-dakota-nuclear-missile-accident-released/ As for the Damascus incident, there's Eric Schlosser's whole book, Command & Control @scruss @johnefrancis I heartily recommend the book Command and Control. @kenshirriff @johnefrancis agreed. The whole thing starts with "You did *what* on top of two tanks of hypergolic fuel?" I think the moral of both stories is : go back to the damn truck to get the right tool  @kenshirriff was the guidance ever upgraded or are those 400 Minutemen still running on this vintage technology? @root42  @kenshirriff That thing about separating the rocket from the ground connection - in a lot of early NASA rocket launches there were incidents (explosions) due to the wield way the launch control/abort circuitry was set up, “sneak circuits” would occur when the rocket was taking off, like just as the rocket disconnected from the grounding plug at the bottom. I’m wondering if that instant severing of connections with the squib might’ve also been to prevent sneak circuits. @DeltaWye That makes sense. Weird things could happen right at launch. @kenshirriff Document with a great example of a sneak circuit involving the cargo door and landing gear in an aircraft. There are some other examples out there. If the emergency door open switch and the normal door open switch were operated, the landing gear would deploy. (Not a sneak circuit but there was an incredible wiring error in a train and a design flaw where sounding a whistle on a locomotive caused the carriages to decouple.)  @kenshirriff @DeltaWye I'm wondering if this still makes sense with modern computers clocking away at GigHz – would pyrotechnic be "fast enough"? @kb9ens I don't know the internals of the power supply oscillators. It could be a quartz oscillator, but I wouldn't be surprised if it were just R-C, since the timing isn't particularly important. @kenshirriff Woah... sounds like that would also be resilient to transient currents and restoration of time after a transient induced shutdown. As ever clever clever...  @kenshirriff @kb9ens that is an interesting combination of design decisions.   @kenshirriff I'm going to guess running Doom on this computer might be difficult, but it will bring it   @kenshirriff This thread is very interesting! Thanks for the breakdown. I'm curious what it means for a computer to be 24-bit but operate with a 1-bit ALU and 1-bit I/O. What part is 24-bit? A word is 24 bit, but arithmetic operations are made one bit by one bit, beginning by least significant one. We do the same in 10 base. first adding the units then the tens then the hundreds... @tanavit @kenshirriff So if a "register" contained 2**24-1 (all 1s) and then the program added 1 (overflowing), the register would see 24 different states (1 for each bit), and after the first cycle it would have appeared to have subtracted 1 bit, first setting the lowest bit to 0 while the remaining 23 bits were still set to 1. Do I have that right? If that is the case, was the program written to include some check-pointing logic to avoid reading interstitial states after a power loss? @dvogel You can think of it as a 24-bit architecture with a 1-bit implementation. Similar to the PDP-8/S, which implemented the 12-bit PDP-8 architecture with serial circuitry. Also, the Datapoint 2200, a serial 8-bit computer whose architecture was copied to create the Intel 8008 chip. (That's why x86 is little-endian; you have to start with the lowest bit on a serial machine.) |
The computer had no RAM but stored everything (code, data, registers) on a hard disk that held just 2727 words. The disk had multiple fixed heads (see photo), so it didn't need to seek. The disk provided one bit at a time to the serial computer, which would perform the instruction and write the results back to disk one bit at a time. 8/N