Recalling the AUTODIN – Part I

Recalling the AUTODIN – Part I

February 24, 2012 4:36 pm 16 comments

Autodin (short for “AUTOmatic DIgital Network”) was the Department of Defense’s first computerized message switching system. It was developed to replace a semiautomatic teletype switching system called Plan 55. Like Plan 55, Autodin was developed in the late 50′s, early 60′s by Western Union. The original five Continental US (CONUS) sites were operated jointly by the military and Western Union. Subsequent sites located in other parts of the world were operated by Philco-Ford and military agencies. Autodin remained in service into the 2000′s and has been replaced by the Defense Message System.

I worked as a tech controller in the Autodin AESC (automatic electronic switching center) at Norton AFB, San Bernardino, California, from mid-1966 until 1968. The Norton switch was the first one to come on-line (around 1962, I believe) and maintained a heritage of being the most reliable switch in the system. The last full year I worked there, the switch maintained an uptime of 99.97%. That may not seem amazing, but you have to understand that this was with second-generation computers.

Second Generation Computers

Autodin Computer Room

Autodin Computer Room

Autodin was initially implemented using second generation computers. These were built using semiconductors (as opposed to vacuum tubes in the first generation) but no integrated circuits. Even simple flip-flops were constructed using discrete transistors, resistors, and capacitors. Usually the circuits were built on small printed circuit boards that were plugged into racks with interconnecting wiring on the back. A large computer typically had thousands of such cards.

Communications Data Processor (CDP)

The primary switching computer for Autodin was called, simply, the CDP. It was a monstrous thing, comprising 19 racks of equipment, each containing hundreds of circuit cards. It was built by RCA and, if I recall correctly, was based on the RCA 501 commercial computer. It was a word-oriented (as opposed to today’s byte-oriented) machine, with 56 bits per word and ferrite core main memory storing 40,960 words. DC power requirements were quite high (hundreds of amperes) and were supplied by large motor-generator sets (AC motors driving directly connected DC generators). It needed a large air conditioner to keep it cool.

CDP Console Layout

CDP Console Layout

There was a wide data bus connecting all the racks together via large cables. Because the signal path was so long, data placed on the bus was parity protected. In fact, the most frequent failure mode was a BTPE (pronounced buhTEEpee), a bus transmission parity error.

It supported a variety of I/O devices. Magnetic drum storage units holding 322 kilobytes were used to store temporary operating data. Console printing was done on Flexowriter automatic typewriters. There was a high-speed line printer for dumps and other large printouts. A bank of half-inch reel-to-reel tape drives provided secondary storage for initial program loading, logging, and off-line message storage. It also had interfaces to multiple Accumulation and Distribution Units (ADUs – see below) and a specialized computer called the Automatic Display Processor (ADP – see below) that drove the system console.

Each switching center had two CDPs, arranged in a hot standby configuration. If the online CDP failed, all I/O devices were automatically switched to the standby processor, which took over with little discernible interruption.

The CDP was notable in the history of computer development because it was microprogrammed. Each instruction in the custom language developed for the Autodin application decoded into a subroutine of simpler operations that the hardware could perform.

The message switching program on the CDP ran in a cyclic manner through receiving, switching, and sending phases. This could be observed easily on the CDP maintenance console, which provided light displays for many of the CPU registers. They were arranged to display 27-bit half words in a 3 by 9 arrangement. Each column of three binary bits could easily be read off as an octal digit, which was used for troubleshooting and debugging purposes. The rapid cyclic flashing of the register displays became very familiar to system operators and usually the first indication that anything was going wrong was a change in the light pattern as the online unit executed a transfer to the standby unit.

Accumulation and Distribution Unit (ADU)

ADUs were the equivalent of what would later be called communication front-end processors. They provided the interface to the data communication circuits connecting the switching center to tributary terminals and other switching centers. Each switching center had three ADUs, with two required for normal operations. The third was maintained in warm standby and could be substituted quickly for one that had failed.

Automatic Display Processor (ADP)

The ADP was a curious little computer used only to drive the System Console, which provided convenient display and control capabilities for operating the message switching system. It was curious in that it used mercury delay lines for main memory. That is, the contents of main memory were kept circulating in a loop through the delay lines, with the processor executing its cyclic program as the instructions streamed by. It was also curious in that its initial program load was from punched paper tape.

Tape Search Unit (TSU)

Tape Search Unit

Tape Search Unit

There was also a stand-alone computer called the Tape Search Unit, which was used to search history tapes to recover messages that had failed to be delivered for some reason. This was the first computer I ever worked with that required manual entry of a bootstrap program to get it started. If the power to the TSU had been turned off you had to manually enter about thirty instructions in binary by punching buttons on the TSU front panel. That bootstrap program could then read the tape search program in from mag tape.

Terminal Equipment

Synchronous Terminals

Compound Terminal

Compound Terminal

The primary terminal equipment used at Autodin tributary stations was called the Compound Terminal (CT) or, more familiarly as “the cube.” It was called that because the processor unit, made by IBM, was exactly square, about 3-1/2 feet on the sides, and about 5 feet high. It provided facilities for transmission and reception of 80-column punched cards using an automated version of the IBM 026 keypunch. It also could send and receive text messages using a Teletype ASR28, which had a paper tape punch, tape reader, and page printer. Messages were prepared by punching them into either cards or tape, then reading them into the system. The CT then transmitted them into the Autodin switch.

Larger facilities used a faster terminal built by IBM, which could punch paper tape so fast that it would fly four or five feet out of the punch before dropping to the floor. Later there was another high speed terminal built by Univac, called the 1004/DLT6, basically a Univac 1004 computer with special data com equipment (the Data Link Terminal model 6) attached.

All the originally planned terminals operated synchronously using (originally) a six-bit character code called Fieldata. Later, the whole system was converted to use ASCII.

Mode 5 Terminals

In an attempt to reduce the cost of Autodin terminal equipment, while still obtaining its automatic acknowledgement of messages by the receiving station, the DOD commissioned Western Union to provide an asynchronous Autodin terminal that could be used with existing Baudot (five-bit code) teletype equipment. The result was an asynchronous system called Mode 5 (the main synchronous terminals were Mode 2).

A Mode 5 control unit could be connected to an existing teletypewriter terminal to provide automatic numbering of outgoing messages, and automatic acknowledgement of incoming messages. With no control characters available in the Baudot code, it transferred control sequences in a novel way–one that was later used with the Smartcom modem control codes to switch from data mode to command mode–the use of idle blind periods. A control sequence consisted of a pair of identical characters that followed an extended idle period on the line. The control unit inserted pauses in transmission so it could send acknowledgements back for received messages, and it automatically detected pauses and control sequences sent by the Autodin switch.

Mode 5 transmissions were not error protected, but there was at least assurance of delivery of each message by the automatic numbering and acknowledgement system.

Continue to Part II >

16 Comments

  • Joseph Reger

    I worked in the Norton Autodin\Crypto shop 1976 to 1979. There were KG-26C, and KG-13 in that time. What crypto equipment was there in 1966 to 1968? I ended up as the senior 2 stripper for over a year as shift supervisor, minum requirement was a 4 stripper. Work week consisted of: 3 days on, 3 days off, 12 hour shifts. They rotated also. Presently I supervise an Open Computer lab with 9 employees.

  • Crypto in the 1966-68 timeframe was KW-26s for the teletype circuits, and KG-3/13s for the higher speed data circuits. I’d have thought that ten years down the road they would have come up with something cryptographically stronger, but I guess not. The KG-3/13 with its completely randomized initialization vector was a godsend for us tech controllers. All we had to do to resync a data circuit was push a button on the tech control console. No need to call into crypto for a card change, as with the teletype circuits.

  • I’ve just been reading these pages and brings back a lot of memories for me. I also worked at Norton AFB. I went to work there right out of H/S and was trained as a teletype operator. Our station was at top of a hanger out near the runway. 3 people on duty 24-7. Below us in the hanger, was maint. going on for Atlas,Thor,missles. In the same room with us were Key Punch operators. After about a year, we were trained to use the new Autodin Compound Terminal. Only one bomb scare where we had to evacuate the hanger. We were then moved down to a large building, where we worked alongside AF men. Those were fun times, so long forgotten. Thanks for some memories.

  • P.S
    Forgot to mention that I was standing at the incoming msg mach. when over United Press came the report that John F. Kennedy had been shot in Dallas.

  • Gerald Joiner

    I was stationed at the Coltano ASC 1971 to 1973, Fort Detrick ECTC 1973 to 1975, Pirmasens 1975 to 1978 and back to Ft Detrick 1978 to 1980 when I left the Army. A lot of good memories, good friends – too bad we lost contact – would love to know what everyone is doing these days. I’ve been medically retired since 2003.

    • Mark Furman

      Was stationed at Ft. Detrick ECTC Fall 71 to Feb 73. Then to Germany, and ETS’ed. Transferred to Detrick from Richie/ANMCC radio room under the Hill, first to USACEEIA, then next door to ECTC/USASTRATCOM.Multiple transfers show in my DD214. Lived in the barracks in between the USASTRATCOM and the USACEEIA buildings.Did some TDY at Meade putting up new NSA antenna field, spring of 72. Can’t find single picture of exterior of ECTC from back then; want it so can show where I worked. Got there as Pfc, made Spc4 there. Did the Tues/Thurs security runs to DC – 9AM, 3PM with locked pouch after typing pouch routing sheet in USACEEIA office. Had a good time at Detrick. Write me at my yahoo email.

  • Gerald Joiner

    if anyone remembers me, would love to hear from you…

  • Stephen Ames, SMSgt (Ret)

    KW-26 gear has been around a long time. I decommissioned a set in the SSO comm center on Kadena in the late 80s. They were a challenge and most of us from that day knew how to reinitiate a card that has been cut without having to “burn” a new one. In my day I had to certify monthly on KL-7s offline encryption gear. http://en.wikipedia.org/wiki/KL-7 The KW-7s were also a pain when they had the 30 wires to configure. [used on Shu Lin Kou, Taipei] They were later retrofitted with card readers. [used on Kadena] I’ve worked with the whole gauntlent of crypto gear culminating with the KG-81/83/84 family when I retired in 1998. I cut my secure comm teeth on the old Model 19s, manual keyboard action and all. Univac 1004 and DCT 9000 and General Dynamics D.S.T.E.. Model 28s! It’s all coming back. Never fades.

  • KL-7′s were still in use in 1969 when I was stationed in Samsun, Turkey. Although it wasn’t my job, I always set up a KL-7 for the new day’s key since the teletype operators were very busy at that time of the day. As I recall, we had two and alternated them, so you had one to decrypt anything received using the previous day’s key. They were only used if all our links were out for some reason and they had to read messages over the telephone.

  • Art Brown

    I worked at the Communications Computer Programming Center (Det 10, 1800 Support Squadron, AFCS at Tinker AFB,OK) from 1967 to 1970 on the Air Force Logistics Systems Autodin System known as, I believe, Set 8 or RAIDS (Real-time Autodin Interface and Distribution System). We programmed Univac 418-11 computers for the AFLC bases (Tinker, Wright-Pat, Kelly, McClellan, Warner-Robins, etc.). Our commanders during my time were LTC Anna and LTC Love. Our offices were at the Universe Building offbase but near Tinker. I was the Project Leader for the installation at Kelly AFB and spent a lot of TDY time there. I also went to Wright-Patt, Warner-Robins, McClellan and, of course Tinker. In addition to the Univac 418-II programming group, we had programmers for the IBM 360′s, Univac 1004s, and had a remote group in Hawaii developing a so-called Pacific Interim Access and Control System using some old IBM equipment. We had about 170 on-site programmers at locations around the world as well as in the programming group at Tinker.
    Your web site brought back a lot of memories for me. It’s amazing how much of the technical details I had forgotten so I am enjoying reading the various parts of your web site.
    Art Brown

  • Howard A. Johnson

    Great article Phil! I started my ASC career at McClellan AFB, working first in the dual homed minor switch (RUVM), then moving over to a mode 5 trib station (EPA) with REAL Western Union equipment. After that I was lucky to move over to the McClellan ASC (RUWM) for a year. My last AUTODIN assignment was at Elmendorf AFB (AJA), in AK. I so loved working AUTODIN and despise it’s replacement, DMS, which I was forced to support marginally at McGuire AFB, NJ! I did get to go back to messaging when I was in Mainz-Kastel, Germany working at first on a Newdealer terminal, then it’s replacement, AMHS which I enjoyed. But these were not AUTODIN jobs, they hung off the remnants of AUDODIN that was being kept alive by different agencies and departments who refused to switch to DMS.

  • Randall Morrell

    I worked in the Primary Criticomm Relay Center at Chicksands, Kelly, Incirlik, Turkey. Maintained all tty and offline crypto available at that time. Favorite was Chicksands. Mode 5 was a favorite in Korea. KL-7′s were a bitch. :lol:

  • Richard Riffe

    Great article. My grandfather is Richard “Dick” Crumm. One of the creators of DIN and later AUTODIN. Unfortunately he passed away in 1985, and I was born in 1986. I dont kniw to much about his work. I’d be interested to know if you know anything about him.

  • Hi Phil;

    I was an Air Force Tech Controller at McClellan AUTODIN (north side of Sacramento) in 1975 and 1976.

    Having signed-up for a 4 year hitch in the AF I had 3 years out-in-the-field (after finishing tech school at Keesler and waiting for clearances to come through) with my duty at McClellan coming after I’d spent 18 months at Karamursel (in NW Turkey)

    Once my year-and-a-half at McClellan was done it was back to civilian life for me (the 4 years in the AF was good as it allowed time for the motivation to grow such that I finished up my EE degree once I
    put away Air Force Blue).

    I could be wrong but I have long held the opinion that being in an “imbalanced AFSC” (Tech Control with an SSIR) career field (they needed us overseas but not really stateside), plus the DOD reg (can’t remember the number) that said you couldn’t be kept over seas all-the-time, meant that the AF had to ‘find’ positions in the U.S. that required SSIR clearances. In the mid 1970s Sacramento was a nice place to be (the weather was similar to central Texas where I’d grown-up) and with the Sierras less than an hour’s drive to-the-east, and San Francisco about a 90 minute drive west, it made for a good tour (though the summer weather could be a bit harsh).

    Regards;

    Paul

  • Marvin E Woody Sr

    I was a UNIVAC tech. I installed UNIVAC 1004 AUTODIN systems at Langley AFB, Ft. Bragg and a couple other posts. Mid sixties.

  • Bruce Morley

    I will throw my 2 cents worth in here. My last duty assignment was at Gentile A.F.S. in Kettering Ohio. Gentile was a part of the Defense Electronics Supply Center, Just seven miles from Wright Pat. I was in Tech Control in their Autodin center. I had just finished 1.5 years at Clark in the PI and worked in primary control for awhile and then up at the FM Hill microwave site. My point is…..the equipment on the hill was microwave and equipment was older than I was. And then I get to go and work with the state of the art equipment. What a revelation.

Leave a reply