The Autonetics D17B Minuteman I Missile Guidance Computer:
A Brief Description of Efforts at Drexel University

Version 1.30 (24-Jan-19)

Copyright © 2012-2019
Samuel M. Goldwasser
--- All Rights Reserved ---

For contact info, please see the
Sci.Electronics.Repair FAQ Email Links Page.

Reproduction of this document in whole or in part is permitted if both of the following conditions are satisfied:
  1. This notice is included in its entirety at the beginning.
  2. There is no charge except to cover the costs of copying.

Table of Contents

  • Back to D17B Table of Contents.


    The Minuteman I Intercontinental Ballistic Missile (ICBM) was the first of the USA's long range deterrents and was deployed in 1962. When they were decommissioned between 1966 and 1973 to be replaced with something more deadly, the advanced (for its time) Autonetics D17B guidance computers were made available free to educational institutions. Drexel University (Philadelphia, PA) acquired three of these beasts. Having been accepted to Drexel and while still in high school in 1969, I "volunteered" to attempt to get the D17B working before officially starting the undergraduate program in the Fall. I seem to recall that the professor who ran the lab did pay my bus fare. ;-)

    The D17B was a stored program computer whose main memory consisted of a fixed head disk drive with a single one-sided platter with approximately 50 tracks storing a grand total of 2,727 24 bit words each of which could be organized as a single double-precision word or two single precision words of 11 bits each. There were 3 timing bits between the 24 bit words on the disk. In fact, nearly all the working registers were also implemented as serial recirculating loops of 1, 4, 8, or 16 words on the disk with multiple read/write heads per track to read the data, change it if called for by the computations, and write it back. The clock rate was a whopping 345.6 kHz with a basic word time and double precision (24 bit) instruction requiring 78.125 µs or each bit being 2.894 µs. This thing really smoked (hopefully not literally). Modern vanilla-flavored commodity PCs run a million times faster not counting the parallelism present in advanced integrated circuit implementations.

    The CPU had only a very few actual hardware registers. All the processing was done with discrete logic where each part had a serial number, even resistors and capacitors, with everything being polyurethane conformal coated. The PCBs were rather ethetically pleasing to behold, even if they didn't do much. :) A 24 bit flip-flip register occupied an entire roughly 4x8 inch PCB. There were no integrated circuits or core, and solid state memory was unheard of (or way to expensive even for the Military) when the D17B was designed. Now think about this: What bad things could happen with a rotating disk spinning at 6,000 rpm flying on a missile. :)

    Before you laugh at the D17B because of its pathetic performance, one should realize that the use by mere mortals of any digital computer in 1969 was a big deal. Most computers back then occupied entire rooms accessible only to the privilaged few, with air-conditioning and a raised floor. One "interacted" with the computer via punch cards and printouts. The first hobbyist computers like the Altair 8800 would not appear under 1975 and the Apple 1 would not be introduced until 1976. Minicomputers like the PDP-8 were available earlier but had starting prices similar to the sum of several luxury cars and useful configurations could cost more than an up-scale suburban house, so they were not any more accessible to ordinary people. Thus while the D17B had less processing power than a smart light bulb or key fob today, it did have the fundamental capabilities that are currently associated with the term "stored program digital computer".

    Physically, the D17B was a toroid about 3 feet in diameter with all the backplane wiring soldered point-point on the inner cylindrical surface. It sort of resembled a mini-version of a (much later) CRAY supercomputer. One half of the toroid contained the disk and logic PCBs while the other half had the internal DC power supplies and torque and (probably) gyro spin motor drivers for the inertial navigation system. Regrettably, the "good stuff" like the nuke and gyros were removed (the latter being considered something like "strategic controlled technology"). And all the data on the disk was wiped by "being overwritten a minimum of 3 (three) times with random numbers" (as stated on the paperwork that came with each D17B. :( :) Obviously, it wouldn't be good for anyone to be able to copy the code to be used for their own private missile. ;-) But everything else of importance was present as well as a few interesting remnants of the inertial platform including precision angle sensors and torque motors.

    For more details on the D17B architecture and implementation, please refer to the links below. The most concise is probably the one for Wikipedia. The following is my brief hazy recollection from almost 50 years ago of the saga of getting it running, designing and constructing I/O interfaces, and developing support and applications software (sort of). For in-depth information, the links or a Web search will be far more useful. Unfortunately, this was way before the Internet as we know it, so no YouTube videos. And few if any documents or photographs detailing this effort were ever created and none survive that I'm aware of. Too bad we can't go back in time and make some nice videos of the register displays flashing, the paper tape reader spinning, and the typewriter clunking away. ;-) Furthermore, at this point I'm not certain of the names of the other students who worked with me or previously. If anyone reading this can provide more details not already covered from sources, or corrections or additions for which I will credit you appropriately, please contact me via the email link, above.

    Note: The links and photos open in a single new window or tab depending on your browser's settings.

    Acknowledgement: The "favicon" for this page was crunched from the D17B photo at: Minutemanmissile Guidance System: A Tribute to the ICBM Program.

  • Back to D17B Table of Contents.

    Hardware and software development

    Getting the D17B to come alive

    As noted, Drexel acquired three D17Bs. D17B Machine #1 - the only one that was actually ever powered at Drexel - was located in the research laboratory of Professor Peter Herczfeld, who has been on the faculty of Drexel University since 1967. Currently, he is the Lester Kraus Professor of Electrical and Computer Engineering. What's interesting is that Professor Herczfeld was not really an engineer and his area of specialization relates mostly to electromagnetic fields, at that time in microwaves, and now microwave-photonics. So, exactly how he ended up with the D17B project was never really clear. Applications of the D17B in microwaves would seem unlikely. However, Professor Herczfeld was also may faculty adviser while at Drexel so perhaps hosting the D17B project was done as a service to the University - or there was an empty corner of his lab where it could be plunked down. ;-) With the grad students doing microwave research (i.e., waveguides and plumbing and stuff) there was little in the way of technical assistance directly available. Anything below 1 GHz was considered to be DC, boring, and had no future. ;-)

    Some previous students had built a nice platform out of steel angle stock and particle board for the D17B to sit at a comfortable height with the power supply below it. Given the extent to which it had been set up, I'd guess these were obtained shortly after the first Minuteman Is were decommissioned, perhaps 1967 or 1968.

    A 28 VDC power supply had already been cobbled together but not much else had been done. I seem to recall that the original DC power supply involved car batteries in some way, shape, or form. Two 12 V batteries in series would be around 24 VDC, which was probably close enough for Government work, but the first thing I did was to replace those with a brute force unregulated 28 VDC power supply so as not to be concerned with available run time as we are now with our mobile devices - transformer, rectifier, lots of filter capacitors. ;-)

    There was just enough documentation to have a chance of being able to develop the needed interfaces and programming the thing. I do not know how much of this was provided by the Government and how much was determined by groups at other educational institutions. However, there was a D17B Users Group and information was exchanged on the finer aspects of D17B hardware and software. There were even several meetings, though I did not attend any. Check out the links below.

    However regardless of the flavor of the DC input, the D17B pretty much just sat there doing nothing. There was little to no activity on any of the logic test-points. And no audible sign of the disk spinning up, though a whine could be heard from the motor driver choppers. That lack of activity was in fact was the first problem since it seems that even the master logic clock is derived from the disk so it can be synced to the data on the disk. With the disk not spinning, no clock, nothing else happens. So why was the disk not turning? Drive voltages were present at the disk connector. Some exploration was in order. With the disk drive removed from its mount someone noticed a warranty seal in the center concealing a bolt. Now what could one do with a bolt? Turn it of course! And, loosening it ever so slightly along with a bit of rotary persuasion to unstick the platter, voila! Suddenly disk spinning sounds could be heard and pulses appeared on those dead test-points! Now, the need to loosen a bolt underneath a warranty seal was probably not a good thing but we were so overjoyed that something was happening. So it did not occur to us that this might have been a sign of impending doom. :( :) It's possible that bolt had to be set to a very precise torque or based on some measure of performance. Or it may simply have secured the bottom to the top and for unknown reasons, the clearances were too close. All we had was a socket wrench and no documentation about that. Lucky there were 2 spare machines just in case. It would be our luck (or lack thereof) that the disk in the first one must have been damaged. My guess would be due to moisture, though it seemed to be well sealed. Thankfully, this one never needed to actually be used in the missile (for more than one reason). Then again perhaps none of these missiles would have actually worked correctly, which was probably just as well.....

    In fact, it was usually necessary to jiggle the bolt to convince the disk to spin up on subsequent power cycles. Eventually signals began to become erratic. But initially, all appeared well based on monitoring various test-points and applying selected logic signals to inputs like RESET. It was doing infinitely more than it had been previously.

    Hardware problems

    Given that the D17B was designed and manufactured to military standards, most of the hardware was reliable and gave us no problems. However, that pesky disk - at least the disk in machine #1 - started failing almost as soon as it was revived. So strange things would happen whereby the serial output streams from various registers or memory locations would simply go dead never to be seen again. And it became increasingly necessary to fiddle with the magic bolt to get the disk to start. Eventually, enough was enough, and the disk from machine #2 was transplanted into machine #1. That one performed flawlessly and remained in place for the duration. So we were just really unlucky in happening to select the wrong D17B to start with. I doubt it was caused by anything the previous group did. I don't recall if the replacement preceeded the software development or was forced on us when reliability declined to the level of uselessness. A post mortum on the first disk - which required quite a bit or pursuasion to get apart including the use of large hammer in leu of an 8 inch spanner wrench - determined that there was almost no magnetic coating remaining on the disk platter. So one final head crash and there was nothing left. That's the disk drive in the photos below. However, the heads did't appear to be particularly abused or clogged and I don't think any particular effort was made to clean it up. The platter itself was furthre abused by standing in for a hockey puck, and vanished long ago.

    At least two of the disk drives as well as most of the PCBs that had been scavanged have since either been sold off privately or on eBay.

    Data input/output

    Two forms of input were implemented: Keyboard input from a CDC paper tape reader and a modified electric typewriter, sort of a like a clunky TTY, probably also CDC). Printed output was to the CDC typewriter thing. There was also a companion paper tape punch, though I don't recall actually doing anything with that. All of these peripherals were present when I arrived so I don't know if they also came along with the D17Bs or were acquired based on recommendations from or via the D17B Users Group.

    Since the D17B did not have any sort of display console, it was absolutely essential to provide something with lights and switches for it to qualify as a digital system. ;-) So a dual register display was constructed (see the photo below) which consisted of two banks of 28 V incandescent lamps so that a pair of 24 bit registers could be displayed in real time. As I recall, there was some built-in way in the D17B of selecting which CPU registers or disk words would appear in the display. I believe the interface from the D17B was parallel as the only active devices in the display were germanium transistor drivers for each lamp. A banana jack was provided for each bit so that it be monitored or used to trigger other devices, though I don't believe anything was ever done with them. The colors of the jacks were based on what was available. ;-) There was a similar smaller 8 bit display and monitor panel built which served the paper tape reader (and possibly the typewriter). Unfortunately, the lenses on that one were salvaged a long time ago for other projects since they were deemed to be the only useful parts, thus no photos. :) The only switches on either of these were red pushbuttons for reset or something, but no switches for individual bits. So, to manually enter data, the typewriter was the only option. Both displays have been sold recently on eBay to D17B aficionados. :)

    Software development

    While it was possible in principle to enter programs in binary via the typewriter, something more sophisticated was required to make this thing at all useful. So a software assembler was written in the style of the PDP-8 assembler in PDP-8 assembly language on a PDP-12 (got that?). All software development for the D17B was done on the PDP-12 (a version of the PDP-8 mated with a LINC computer for interfacing to lab experiments) in Drexel's Bioengineering Department. Those in charge there probably weren't real happy about us using it for this stuff but tolerated it, probably holding out hope that a (1) the D17B might end up being useful and (2) one might be made available for their experiments - or something. I do recall that it was best to use the PDP-12 during off-hours to minimize disruption of funded research activities.

    The D17B assembler had most of the basic features of the PDP-8 assembler but with the addition of optimization based on the location of the data on the disk. While the D17B hardware would perform correctly regardless of how memory was organized, that could result in totally pathetic performance due to disk latency. That is, even more pathetic performance than imaginable for a serial computer running off a disk! For example, if a program attempted to access two numerically adjacent words (on the disk) in successive instructions, an entire revolution of the disk would be required to fetch the second one because it would have passed over the read head by the time it was needed. The numeric results would still be correct - if one waited long enough. The assembler would auto-magically space out memory addresses to (more or less) minimize such issues. Or at least that was the objective

    Code was entered into the PDP-12 via an ASR-33 TTY and stored on DECtape (those funny runt-size reels with a short length of directly accessible magnetic tape). Binaries were output on paper tape for input to the D17B. The entire process required a trip to a different building to modify the program using the PDP-12, print out the updated program, punch a paper tape, then return to see if what was done resulting in anything useful. Not exactly interactive computing. But we had control over the entire process and didn't have to depend on the Burroughs mainframe at Drexel.

    Application software

    The only "App" that I recall actually being developed was a calculator modeled along the lines of the HP-35 (which was new and cool at the time). However, I doubt it had more than the basic 4 functions. And possibly not even divide as that was not a D17B hardware instruction. It used reverse polish entry like the HP-35 with a 4 number stack, using the typewriter for I/O. It may have even been on in HEX, not even decimal. ;-) But it proved out the functionality of the hardware and the assembler. And I was definitely not a software type back then (nor am I now though there have been transgressions into modern Arduino and MIPS coding and such).


    I took these photos many years later and posted those of the PCBs and disk drive to Wikipedia, but they were removed for unknown reasons.


    -- end V1.30 --