NTDS UNIT COMPUTER
    (a.k.a. UNIVAC CP-642, 
    UNIVAC 1206, and AN USQ-20) 
    Remington-Rand Univac Division


    Word Length 30 bits

    Speed: 9.6 microseconds add time.

    Primary Memory: 32,768 words core memory (3.6 microseconds access time)

    Secondary  Memory: Magnetic drum and magnetic tapes,

    Instruction Set: 62 30 bit, single address instructions.

    Architecture: Parallel, binary, fixed point arithmetic, 7 index 
    registers, 1  accumulator register, 1 free register. 

    Technology: 10,702 transistors,

    Input and Output: Punched  cards,  paper tape, CRT

    Price: $500,000.

    Size: 58.6 cubic feet, 2,320 pounds, 25 kW

    Software: CS-1 compiler

    Development History: Developed under contract for the Navy 
    Tactical Data System (NTDS) by the Saint Paul division of
    Remington Rand Univac. Seymour Cray was the primary
    logic  and  circuit designer.

    Production History: The first units were  delivered in 1958/
    Later commercially available as the UNIVAC 1206  .

    Use: Real-time tactical analysis, display and control of weapons.       
    

The following images are from "Duane"

• Unisys History Newsletter also note UNIVAC M-460
• a USS Nimitz Photo Collection by Dan Luper

Received from Lyle J. Franklin - Mar 2008

I was on staff to Vern Leas, NTDS Program Manager, when he made the decision to drop the 17 which was designed by Seymour Cray. At decision time Seymour had already left for CDC and was not involved in the design of the 20. The placord should be corrected. People I remember involved in the design were Finley McLeod, Hy Osofsky, Glen Kregness and many more. Curt Christensen was involved in the decision to redesign. As I remember it, we built serial sixteen first to serve as the prototype. The 17 was supposed to be the NTDS prototypes. Curt has tried to correct many of the errors appearing in the historical documents. Curt can be reached at 651/457-1401. Lyle J. Franklin Life Number 5885

Received from Jack C. Lamb - Mar 2008 - 'Battle-Wagon' and 'StarTrek'

I was a lowly, button-mashing twidget (OS) in the US Navy circa late 70's/early 80's and, when we were moored and had a stable power supply from shore, we always tried to talk the DS's into loading up 2 great games that were played in CIC on the NTDS consoles... I really liked 'Battle-Wagon' and the other was 'Star Trek' or similar. Would you know anything about the history of these early games (i.e. who slipped these to the fleet and who programmed these on the side?) I doubt the Navy contract for the AN/USQ-17 or AN/USQ-20 'officially' included these diversions (but they were great in their day!)...it seems no one wants to remember or talk about them... Thanx!...signed OS1 Jack C. Lamb/USS Luce/DDG-38 (R.I.P.)

Received from Barrett, Joshua - May 2007

I was an OS on the Carl Vinson from 1988 to 1991. We still used 3 CP-642B to run the op program. They came from the Oriscany (I was told) - - - - later - - - - they were replaced when the carl vinson went to the yards in late 1990. it is on the chuck boat web site. Thanks! There is still a CP-642B on the midway museum in San Diego in the SINS room. There is also one at the computer history museum near google in Mountainview CA.

Received from Gary Antrobus - Centerville OH - April 2007

What great memories! One of my assignments as a new officer in the Navy in 1968 was to teach programming for the AN/USQ-20, both machine language and the CS-1 compiler. During the machine language part of the training we used Frieden Flexowriters to punch paper tape for program and test data loads. We had to debug using the register panels on the top of the units by executing one instruction at a time and reading the output lamps. I got so used to base 8 math that it took some time to relearn base 10 after I left the assignment. I still have the laminated instruction card. The 8 week course we developed was fantastic. A junior officer or civilian could enter the program with little or even no knowledge of computers and two months later could be relatively productive working on the various NTDS modules at either FCPCLANT or FCPCPAC. I really loved this duty and I got to work with some of the most brilliant engineers and technicians I have ever met. The UNIVAC people and the navy ETs were amazing!

Received from Larry Schneller

I was a Navy DS (Data System tech)from 1961 to 1966. I went to school on the 642A at Mare Island CA. After the Navy I worked for several years at Pt. Mugu where we had a 642 and several 642 Bs. Later they got a 1230 almost the same machine but a few more capablities (developed for NASA). Some of the info in the write ups didn't sound right right. The memeory cycle time (read/write) on the 642 was 16 micro seconds. The CPU had a 4 phase clock (4 micro secs per phase). This info is about 40 years old. I may have an old Tech manual for the Computers some where. The picture at the web site did not have the maintenace panel which sits on top. I stuck with the Univac computers for years and advanced to the UYK 7 in the 70s.

Received from William H. Bennet received Nov. 10, 2002

Hi Ed, I was a DS from 1967 to 1974. I had a 1668 job code for computers and peripherals. The 642A was all germanium transistors. The 642B used silicon. Both had a 1 MHZ clock, divided into four phases. Instructions were from 4 microsec. to 48 microsec., depending on the operation. Memory was ferrite core. Cycle time was 8 microsec., 4 to do the destructive read, and another 4 to write back in from of the Z register. Both had an unusual hardware square root routine. The op code was for divide (23), but it extracted the root if the k register was 7. The X register did an elaborate shift process, which took 48 microseconds. I believe your figure for power consumption is high. We ran 3 642A's, a RD-281 magnetic tape unit, the paper tape reader/punch, and several other pieces of gear off one PU-491 motor-generator, rated at 5 Kw. The power supplies were unregulated, and used 400 Hz 3-phase power to minimize ripple. The fleet had no assembler/compiler, so we had to do our programming by sitting on a tall stool so we could manually enter each instruction into the A register. Each instruction was manually stepped into memory. If nothing else, you learned effecient coding. We wrote some maintence routines, and a few games. I think every DS wrote a patch for War or Spaceship. We also has Baseball and Football. All were played on the Hughes Aircraft radar display consoles. You can't really apply core technology criteria to these old Univacs. In a way, every manufacturer had a core technology, but things were still evolving from some really primitive beginnings, so I wouldn't think of the differences as such. The distinction between a 642A (Univac 1206) and a 642B (Univac 1208) was pretty slight. The biggest difference was the change from germanium to silicon transistors. As the germanuim transistors were non-standard, the change probably save Univac significant cost per unit. The old germanium transistors were housed in a grain of wheat sized casing, with fine insulated wired coming out one end. The transistor had to be mounted in a nylon clip, and then each wire had to be stripped and hand inserted into its hole in the pc board. Has to be labor-intensive. Both computers used PNP transistors, so a logic Low was true, and a High was false.TTL was hard to get used to after I got out of the Service. The other differences were small. The 642A had more elaborate circuitry for memory. There were manual adjustments for the X, Y and Z drive currents. The 642B had a Z current adjustment only, as I recall. The 642B could have its I/O channels sped up from 125 Kword/sec. to double that rate. Some problems apparently crept in, as there was an engineering kludge in the ODR timing circuit. It was two flip-flops that had been added. One was named Finkbine, and the other Gomez. I noticed that the newer Mil-Standard 802C prints omitted the names. It just popped into my head that the logic in the 642A was basically all NOR gates, while the 642B was all NAND gates. The prints were reprenented in something called bubble logic, where each element was a circle. 802C was lots easier to interpret. Without prints and such, that's about all I can recall of these computers. Hope this helps, Bill Bennett