The Univac 1 Computer

by George Michael

Figure A - The LLNL Univac Console Area	Figure B - The Acoustic Memory Units
This image shows the console of the UNIVAC 1 and some of the ten UNISERVOs (tape units) in the background. Lou Linnager is using the toggle switches at the console, while Earl Means is hanging a tape. On the right is a Remmington Rand typewriter, suitably modified to serve as the on-line printer for the system; there was also an off-line typewriter-printer that used output tapes written at 20 cpi. This machine is usually regarded as the first commercially available computer in the U.S. Ironically, its development was supported largely by a commercial enterprise - the Franklin Life Insurance Company (apparently up to the 1940s, there were businessmen with a modicum of vision). The UNIVAC 1 had only 1,000 words of memory, each word containing 12 decimal digits, and each digit being 7 binary bits, counting parity. This was called the "Excess 3" code, meaning that 3 was added to each digit, so that zero was coded as 3: `0` (decimal) = `100 0011` (binary) The leading "1" is the parity bit, ensuring odd parity. [1] The magnetic tapes were actually metal tape coated with a magnetic oxide. High density was 200 cpi, and as noted above, 20 cpi for both input and output.	I don't know the pictured technician; I don't remember him at the Lab, so he may have been a UNIVAC employee. (Anyone who does, please let us know.) Pictured here are three of the ten memory units. All memory was housed in a roughly 10' by 8' by 6' walk-in box whose sides contained the UNIVAC components and wiring. The memory consisted of Mercury delay lines, each of which contained 10 words. The data were represented as a serial string of acoustic pulses circulating in a Mercury line. As the pulses emerged, they were reshaped and reinserted into the delay lines - memory! Remember, the UNIVAC was a decimal machine. Each memory unit had 100 words, so there were 1000 words total, and each was 12 decimal digits in length. There were 7 binary bits per digit, including 1 bit for parity, 2 zone bits, for alphabetic coding, and 4 bits for decimal number representation. The number system was "Excess 3" [1]. This, we were told, was to avoid ever having to represent a digit by all zeros, a limitation imposed at the time by the capabilities of the components - tubes (or valves, as the British say) and power supplies. Actually, there were more that 1,000 words in the memory system. Certain buffers could be used (if you were clever) as other memory. One-word, two-word, ten-word, and sixty-word buffered Tanks were provided for input and another full set for output operations. Programmers quickly learned how to use these efficiently. Another design feature of the memory stimulated a programming style called Minimum Latency Coding, which designed program flow so that operands became available just as they were needed.
The Univac 1 was seminal both as the first commercial computer and as a scientific computer. Others have studied it and produced web pages about it. One such website comes to us courtesy of Al Reiter and can be seen at http://mywebpage.netscape.com/reitery2k/univac1.htm.

A Univac Album

Click on a picture to see a larger (but medium sized) version;
click on a file-name link to see a very large version.

The Console

Figure 1: The LLNL Univac Console
./images/Univac.1.big.jpg and ./images/UNIVAC4.jpg
This is an excellent picture of the UNIVAC 1 console. It is easy to see that the operator has to be somewhat of a virtuoso switch toggler. The UNITYPER provided the remarks that the machine wanted to say to the operator. In the jargon of today, this computer was used 24/7. Generally, at least two persons were always present: an engineer/operator and a programmer.

Figure 2: Another Univac System Console
./images/Univac.jpg

Figure 3: The Programming Switches
./images/Scan6.jpg
This is a portion of the UNIVAC console, used by both the engineers and the programmers. Suitably informed, every aspect of the computer could be observed on this console, and generally used to clear errors and data not expected.

Figure 4: Operators at the Console
./images/Picture.3.jpg
Left to right, Dave Rogers, Barbara Schell and Earl Means. The photograph shows a good view of the UNIVAC console, and the array of magnetic tapes. In the right foreground is the console typewriter, a modified Remington Rand typerwriter.

Figure 5: The Console and Frame
./images/Picture.4.jpg
This is a view that shows the UNIVAC frame, a hollow box. It was a shell with doors that could be opened to expose the computer's compoments, and inside where the memories were placed.

Figure 6: Univac Operator's Console and Operator
./images/UnivacOperatorConsole.jpg
This is a closer view of the Univac Operator's Console. The Operator is Joneal Williams-Daw.

Hardware

Figure 6: Inside the Frame
./images/Picture.5.jpg
This picture shows the inside of the UNIVAC frame during assembly. As noted in Picture 4, the sides of the frame contained all the components of the central computer.

Figure 7: Memory Units in the Frame
UnivacDelayLineMemory.big.jpg

Figure 8: A Chassis Board and Tube-Based Circuitry
./images/Scan10.jpg
This shows a typical set of pluggable chassis that containeds scads of 25L6 vacuum tubes and Gemanium diodes. The strange looking devices on the left are grasshopper fuses, the main protection against bad voltage swings. One night, lightning struck the main transformer situated outside B100. Before the fuses could do their job of protecting the UNIVAC 1, practically every tube in the machine was blown. We (cleverly) concluded that Nature was faster than our UNIVAC. It took almost two weeks to find and replace the damaged components. Steps were taken to ensure that that would never happen again, and sure enough, it didn't.

Figure 9: Peripheral Circuit Components
./images/Scan11.jpg
This shows some components used to make the power fed to the UNIVAC 1 UNISERVOS (tape drives) safe and otherwise acceptable. [2]

Figure 10: A Vacuum Tube Ager
./images/UNIVAC5.jpg
This is a "(Vacuum) Tube Cooker"(Ager). It turns out that the vacuum tubes as received from the suppliers were not uniformly good enough to survive usage in the UNIVAC environment. The main failure mode was caused by bad filaments/cathodes. This simple device allowed power to be placed on the filaments (mostly 25L6's) so they could be aged for on the order of 200 hours. Many failed to last this long. Those that survived were deemed good enough for use in the UNIVAC. This "Tube Cooker" was designed and built by Dick Karpen, and its use to age the tubes that were placed in the UNIVAC dramatically improved the reliability of the computer.

Figure 11: Magnetic Drum Storage Units
./images/UNIVACMagneticDrums.jpg
Although these drum storage devices were available, our UNIVAC did not use them. We relied instead, on magnetic tapes.

Figure 12
./images/Scan5.jpg

Figure 13
./images/Scan8.jpg

Figure 14
./images/Scan12.jpg

Figure 12, 13, 14: Custom Equipment Built at LLNL
These figures show a piece of homemade electronic test gear built by Lou Nofrey and his group to check whether a given Germanium Diode was good enough to be used in the computer. There were approximately 25,000 diodes in the UNIVAC. I've left these pictures in this gallery to show to those who may have never seen a vacuum tube or the large sizes of such test and control equipment. [2]

Peripherals

Figure 15: Univac Magnetic Tape Drives
./images/MagTapeDrive.jpg
This shows the levers, strings and pulleys that were used in the tape handlers for the UNIVAC 1, the UNISERVOS. To keep them working, we had to have on hand a large supply of good fishing line. (It was a great relief when tape handlers with vacuum columns became available.)

Figure 16: The Tape Handling Mechanism
./images/Picture.6.jpg
The ten tape handlers were known as UNISERVOS. These were designed before tape handlers used vacuum columns to maintain the tape loops. In place of vacuum activated controls, the UNISERVOS used springs, pulleys, and fish line.

Figure 17: A Flatbed Plotter
./images/Benson.Lehner.jpg
This is a view of the graphics plotter, with Chet Kennrich trying to tune the system so it would be useful. We tried initially to use it with input from a UNISERVO, but that failed for several reasons, including system noise and analog errors. The plotter was modified to run from punched cards, but that didn't fix any of the problems, and things took so long to get done that it was better and faster to plot them by hand.

Figure 18: The Plotter
./images/UNIVAC2.jpg
This the infamous, clunky plotting device described elsewhere on this URL. Suffice it to say here that it was hard to use. Although it is not certain, the person leaning on the unit looks like Dick Conn, and Katherine Cochran is running the desk calculator.

The Building, The Room, and The Enclosure

Figure 19: Building 100
./images/Scan4.jpg
This is Building 100, the first and only home of the UNIVAC 1. The Neptune moving van is here caught in the act of delivering the computer. Behind B100 is a wooden building that began its life as a Naval Hospital, and became a programmers' warren when the Lab began operations.

Figure 20: Machinery Delivery
./images/Scan9.jpg
The UNIVAC installation included a brawny air conditioning system, here shown being delivered.

Figure 21: Air Conditioning
./images/Scan2.jpg
This shows a portion of the air cleaning system being installed for the UNIVAC 1. It looks like Bob Crew holding one of the filters.

Figure 22: The Frame, in Assembly
./images/Scan3.jpg
This is a front face view of the UNIVAC 1 box. without its covers. The box is hollow. The whole box measured about 12 feet long by about 8 feet wide and about 8 feet high. The memory tanks (Figure B and Fig. 7) are located inside the box.

Figure 23: The Frame, in Use
./images/Picture.4.jpg
This is a view that shows the UNIVAC frame, a hollow box. It was a shell with doors that could be opened to expose the computer's compoments, and inside where the memories were placed.

Milestones and Celebrations

Figure 24: The One Year Birthday Party
./images/univac.party.jpg
Celebrants in this photograph include, from left to right, Leota Barr, Bob Abbott, Tom Wilder, Carl Schneider, Bob Price, Jim Moore, Larry Harrison, Dick Karpen, in the back, Walter Anderson, a UNITYPER expert from UNIVAC (arm around Cecilia), Cecilia Larsen, Jules Mersel in the back directly behind Cecilia, George Michael, Oscar Palos, Chet Kenrich, Merritt Elmore, directly behind Chet, Earl Means, Dana Warren, unknown, Sid Fernbach, Lou Nofrey, and John Hudson.

Figure 25: Celebrants at the Console
./images/Univac.Party.big.jpg
The logic is compelling. The first computer at the Lab was the UNIVAC 1. It arrived at the Lab in April, 1953. To celebrate some no-longer-remembered special UNIVAC event, Sid and Cecilia Larsen threw a party. To be brutally candid, it all was a bit premature since the Lab had yet to design and field its first successful device. However parties have their own raison d'etre, so it was cakes (but no Ale) and lemonade, and coke, the programmers' aqua vitae, and lots of good vibrations. Inside our government lab, one could design nuclear weapons but don't get caught drinking alcohol. So some of the celebrants merely clustered around the control console, maybe hoping that some stray electrons could substitute for alcohol. Seated is Joneal Williams-Daw, and from the left clockwise are standing Stan Helmici, Ed Lafranchi, Cedric Eastburn, Marvin Lehman, Sid Fernbach, Pierre Noyes, and John Hudson. (Special thanks to Cecilia Larsen for helping me recover these names.)

Figure 26: Cutting the Cake
./images/UNIVAC3.jpg
When you are using your first computer, it is not unreasonable to celebrate its birthdays, in this case the third birthday. The persons in the front row, from the left, are Tom Wilder, Sid Fernbach, and Chet Kenrich. Behind Wilder is Richard von Holdt, Ruth Kilby and Mary Ann Mansigh. I am peeking out from behind von Holdt, and Bob LeLevier is behind Ruth Kilby. I don't recognize any of the others in the picture.

The Four-Year Celebration

Univac
1953		1957
April 10th, 1957 marks the end of 4 years' continuous operation of our Univac #5. These have been intriguing years for those of us directly or indirectly associated with this machine. We have, in this comparatively short period of time, seen the Univac become one of the most important tools on the Livermore Project; then, as newer, special purpose machines were introduced, mellow into its present status as "the good old reliable". Since April 20th, 1953, through February 28th, 1957, our Univac has processed 184 scientific problems which involved the typing of 21,600,000 digits of information. This work was accomplished in approximately 19,130 hours of actual productive time. If mathematicians were to accomplish this same amount of calculations, it would require 440 of them to work 100 years, using desk calculators constantly for 42 hours per week, 52 weeks per year, without vacations, sickness or even coffee breaks -- and making no mistakes!! A great deal of electronic material was used to maintain this high productive efficiency. For example, if all the vacuum tubes replaced into the computer were placed end to end, they would stand as high as 2 3mpire State Buildings, with the Eiffel Tower used as a flag pole. Over 4 million kilowatt hours of electrical energy were used during the 4 years. This is more than the total amount of energy generated by the Grand Coulee, the Bonneville and the Hoover dams. This feat of majestic productivity was the result of much tlme and strenuous effort on the part of many people -- physicists, mathematicians and engineers.

UNIVAC ROSTER
Abbott, Bob Alder, Bernie Anderson, Walter Badger, Clarence Barr, Leota Baumhoff, Lester Bentley, Dick Bjorklund, Frank Campbell, Shirley CanfieId, Gene Carr, Bill Cecil, Alex Chuck, Wong Clow, R. Colgate, Rosie Cralle, Bob Crew, Bob Culler, Glen Cunningham, Leland Curley, Ethe1 Daw, Royal Dewey, Tom Dobler, Champe Dorresteyn, Steve Eastburn, Cedric Eckert, Hazel Efstathiou, John Elmore, Merritt Elsworth, Kent	Ferris, Ervie Frank, Jim Fernbach, Sid Gardner, Doug Gerkin, Bill Gersten, Esther Godshalk, Libby Goldberg, Gene Goldberg, Jean Gonzalez, Viki Grayson, Bill Green, Dick Halliday, Glen Hanerfeld, Harold Harrison, Larry Haussmann, Carl Heims, Steve Henyey, Louis Hopkins, Jerry Hudson, John Ishihara, Terry Johnson, Barbara Karpen, Richard Kenrich, Chet Ketcham, Fred Kilby, Ruth Kilpatrick, James Noh, Bill Kinney, Ken	Kuckuck, Paul Lafranchi, Ed Lasher, Gordon Larsen, Cecilia Lehman, Marvin Leith, Chuck LeLevier, Bob Leshan, Ed Levee, Dic Lewis, Marlene Lyons, Jim Mangelsdorf, Bob Manlove, Spencer Mansigh, Maryann Mersel, Jules Michael, George Miller, Ed Moore, Jim Moore, Ray Neilson, Dave Nofrey, Louis Norton, James Noyes, Pierre Nutting, Tane Oeder, Bob Oldani, Charles Oliver, Jack *Palos, Oscar Patterson, Dan	Pennington, Ralph Price, Bob Quong, Jim Rogers, Dave Rose, Don Rose, Jack Sack, Seymour Scharff, Morris Schneider, Bob Shepperd, Mary Sherman, Nevin Stockwell, Gil Telesky, Ray Tiede, Ken Thomas, Stacy Thompson, Ralph Thompson, Shirley Trulio, Jack von Holdt, Dick Warren, Dana Weber, Helen Wilder, Tom WiIets, Larry Wilkins, Mark Wainwright, Tom Wirsching, Joseph *Wythe, Don Yoshisuka, Jean
* 1953

A Description of Univac

The Pocket Univac Reference Card, front

The Pocket Univac Reference Card, back

The Univac Reference Card, pdf version

Corporate PR

Figure 27: System Brochure
./images/Picture.7.jpg
This is a company-designed fanciful layout showing all thre devices one could hook onto the UNIVAC. Included was a device for preparing tape input, the UNITYPER, a printing unit, the UNIPRINTER, special tape handlers for card (Remington Rand style) to tape usage, and so on. Note the ubiquitous Tektronix oscilloscope. Used to diagnose malfunctions, it was known to sit in the corner muttering to the CPU, "Without me, you are Nothing."

Figure 28: System Brochure Part 1
./images/UnivacBrochurePart1.jpg

Figure 29: System Brochure Part 2
./images/UnivacBrochurePart2.jpg

Figure 30: System Brochure Part 3
./images/UnivacBrochurePart3.jpg

[1] Rick Bensene, a correspondent of this web site, offers the following explanation of the Excess 3 code: "The reason for this numeric representation is that it is easier to build the logic gating that handles things like interdigit carries and borrows when counting or performing addition and subtraction. Back in those days, every logic gate took a significant amount of components to build, and component savings was very important in terms of managing the final cost of the computer. Other codes, such as BCD (binary coded decimal) would have required more complex circuitry to implement. Excess-3 code was determined to be the most efficient way to represent decimal digits in four bits, and minimize the amount of circuitry needed."

[2] I am indebted to Ed Lafranchi and Cecilia Larsen for helping to identify the persons and contents of Figures 9, 12, 13, 14, 25, and 26.