Friedrich W. Kistermann
Abstract. The DEHOMAG D11 tabulator has been overlooked for too long in the history of data processing.
This is due to the scarcity of literature about this machine and its usual classification as a tabulator, even though its internal structure corresponds to an automatic calculator. In general, very little attention has been paid to the time period preceding the electronic age, that is, there has been a neglect of the pre-computer era. The D11 tabulator, however, had a decisive influence on the diffusion of punched card data processing in Germany.
1—
Setting the Stage
The development of the DEHOMAG D11 cannot be understood without looking at the development of the Hollerith tabulating machines in general, and at the differences in the application environment on both sides of the Atlantic in particular.
It is helpful to understand the purpose of a tabulator, which becomes apparent if we look at its position in the data processing work flow. Fig. I is an abstract representation of the processing stages. Whenever someone performs any kind of data processing work, for instance when a database management system is used for some task, the same basic steps which are outlined in Fig. 1 must be followed.
True enough, the steps have different names these days, because punched cards are no longer used, but the work flow is generic and independent of the equipment used. In this example, the punched card contains important data concerning a person, event, or product, that is to be processed by the system. The punched card is, therefore, the unit record required for the specific application, let's say for the computation of gross and net wages.1 For this process to function properly, the individual steps have to be carefully planned.
1 Friedrich W. Kistermann, "The Invention and Development of the Hollerith Punched Card: In Commemoration of the 130th Anniversary of the Birth of Herman
Right from the beginning the data to be processed or just stored on a card must be verified, and corrected if necessary. After sorting, which is the process of grouping data in the predefined order necessary to obtain a specific result, the tabulator can do its work. Normally, it was used to print a report and it often produced summary cards as an important by-product. The summary cards made further processing of the data possible, especially in the case of cumulative processing. Because of this logical sequence, the tabulator stands at the end of the processing chain.
Figure 1
Role of the tabulator in the DP work flow
Hollerith and for the 100th Anniversary of Large Scale Data Processing," Annals of the History of Computing 13 (1991) No. 3, 245–259.
The work flow diagram shows that every punched card installation needed at least one card punch, one verifier, one sorter and one tabulator, assuming that the customer was not contracting a service bureau.
In 1895, Herman Hollerith used his punched card system for cost accounting in the railroad industry. This happened in parallel to his work for the United States census bureau and census bureaus in some other countries.2 In 1908, after approximately ten years of development and a major change in construction, the Hollerith punched card system was ready to be widely used in industry and trade.
The Hollerith Punched Card System
The Hollerith punched card system consisted of a key punch, a gang punch, a vertical sorter and a tabulator.3 The Hollerith tabulator of 1908 formed the basis for the successful introduction of the punched card into many application areas.4 The card feeding and sensing device were on the left of this nonprinting tabulator. The plugboard, which allowed the machine operation to be defined by wiring, was on the front. The electrical impulses went from the card sensing brushes to the hubs of the plugboard, and from there by wire connection to a specific counter, where the amount punched into the card was added. The counters, with nine digits each, were located above the plugboard and were visible from the front.
The tabulator underwent some further development after its initial introduction in 1908. In 1911, a manually operated clutch was introduced, which allowed the counters to disengage from the counter clearing shaft and thus retain a result until it was recorded. This created the possibility of two-level group control. The groups were separated by so-called stop cards, which were not punched. If the read brushes detected a stop card, the tabulator would stop feeding cards. The results were read from the counter wheels and were transcribed onto prepared forms. Thereafter, the operator had to clear the counters and restart the machine.
2 Friedrich W. Kistermann, "Ein Kapitel Industriegeschichte: Herman Hollerith (1860–1929) - Ein Pfälzer Emigrantensohn gründet einen Weltkonzern," in Nach-bar Amerika: Verwandte – Feinde – Freunde in drei Jahrhunderten, ed. Gudrun Schäfer (Landau, 1996) 233–267.; and Friedrich W. Kistermann, "Was the Father of Herman Hollerith a Revolutionary?," Annals of the History of Computing 19:4 (1997) 69–70.
3 Friedrich W. Kistermann, "The Way to the First Automatic Sequence-Controlled Printing Calculator: The 1935 DEHOMAG D11 Tabulator," Annals of the History of Computing 17:2 (1995) 33–49; and Friedrich W. Kistermann, "Locating the Victims: The Nonrole of Punched Card Technology and Census Work," Annals of the History of Computing 19:2 (1997) 31–45.
4 Kistermann, The Way, Fig. 3.
Hollerith's Printing Tabulator
The next step in the development of the Hollerith system was to add a printing device. This was also invented by Herman Hollerith. He was granted the patents in 1912, although he had formulated his ideas in 1899. The engineering model was ready in 1917.5 The printing device was attached to the nonprinting tabulator on a pedestal extending to the right of the counters. In this model, the stop cards were replaced by an automatic group control. Two sets of card sensing brushes allowed two successive cards to be compared. If the group identifiers were not equal, the tabulator stopped feeding cards, printed the counters' contents, cleared the counters if they were coupled with the clearing shaft, whereafter the machine resumed card feeding. This provided an uninterrupted work flow at the printing tabulator.
The ability to print out the contents of every card was especially advantageous. It was necessary when statements of accounts, inventory lists, detailed sales analysis reports, etc., had to be processed.
With every new function on the tabulator, the plugboard increased "in size," that is, more hubs had to be wired.
The Hollerith printing tabulator, announced at the end of 1920, marked the transition of the tabulator from a
"statistical" machine into a machine that could support business management in many more applications.
However, there was still one shortcoming in the 1920 model. Negative results were shown at the counters as complementary numbers and were printed as such. A digit 9 in the highest counter position indicated a negative result. To do subtraction, the operators also had to punch complementary numbers into the cards.
Table 1: Application environment: Germany (DEHOMAG) vs. USA (IBM)
We now come to the second half of the 1920s, which saw an increase in demand on the part of existing and prospective customers for more functions in the tabulator and more capacity on the punched card. There was a particular need for a balancing feature, which would allow account balances to be obtained, not only in the banking business, but also in factory accounting and other applications.
The DEHOMAG engineers in Germany became aware of the fact that the Hollerith/IBM tabulators had some deficiencies with regard to application environments in that country, which were quite different from those in the United States. Table 1 contains some examples of the differences between applications in Germany and the USA. The importance of the application is given by the number of stars, from one to three.
Banks wanted to be able to print out complete bank account statements which would show all transactions. In other words, they disliked having one form for credits, another for debits, and a third, which was handwritten, for the balance. Accounting in industry and trade had the same requirements. Cross-footing would be the answer.
If an account has a variable number of entries, which is usually the case, and the balance is to be printed in the last line of the form, then some type of forms control is needed. Also, banks wanted to print true numbers, marked by an appropriate sign, when a negative balance was present. They also wanted to avoid punching complementary numbers instead of negative values. And last, but not least, they wanted to avoid the burden of manual interest calculation by having the tabulator calculate compound interest.
The second line in the application environment table shows the demand on the part of large companies for help in wage accounting. This was due to the more complicated tax reduction procedure that existed in Germany at the time compared to the system in the United States. Cross-footing would also be the answer to this problem.
In wage accounting, cross-footing is cross-addition of all deductions and all additional charges to obtain two sums, which are added with the corresponding sign to the gross wage, printing all details and partial sums on one pay slip.
Some of these requirements were met by introducing additional features in the Hollerith/IBM tabulators then installed. This work was done by DEHOMAG engineers and technicians under the leadership of Ulrich Kölm, who was the most prolific inventor of the group.6 Details of his work cannot be given here, due to space limitations.7
Since the plugboard of the 1920 tabulator could not be extended, the engineers had to add a number of smaller plugboards, placing them just beneath the printing device.8
6 Kistermann, The Way, Fig. 6.
7 Kistermann, The Way, 43.
8 Kistermann, The Way, Fig. 5.
The period between 1926 and 1931 is very interesting as far as the inventions at DEHOMAG in Berlin are concerned. When the summary punch was added to this tabulator in 1931, another small plugboard had to be added. However, the wiring of so many plugboards became something of a problem.
The important improvements in these upgraded machines were:
• the balancing device
• introduction of 9's-complement instead of the 10's complement arithmetic which had been used since Hollerith's time
• introduction of 9's-complement arithmetic for punching negative values
This is remarkable because it made the punching of negative numbers easier and less error-prone. From then on, negative numbers were flagged using an X-punch in the highest column of the card field.
To accomplish this distinction, selectors were added to the tabulator. Selectors are multicontact relays which allow the transfer of negative data to another counter, for example, thereby separating them from positive data when the responsible selector is engaged by the X-punch. That is, selectors are used especially for control purposes. This "field upgrade" could be installed in the tabulators at the customer's site.
Although it was not possible to install the cross-footing feature in these machines because that would have demanded a fundamental change in the construction of the tabulator, DEHOMAG's customers knew that cross-footing was possible in principle. The book-keeping machines of that time, such as the Elliott-Fisher, the Moon-Hopkins or the National, all had this feature.
These developments, coupled with customer demands, caused the DEHOMAG engineers to think about designing a new tabulator, independent of the IBM Corporation. The reasons for this were varied, but the driving forces were, firstly, trying to overcome the wiring intricacies of the enhanced IBM IIIA tabulator and, secondly, the addition of the cross-footing feature. At that time, the IBM technicians were in the process of developing a universal tabulator, a multiplier, and were struggling with the alphabetic printing feature, not to mention the introduction of the 80-column punched card in 1928. Therefore, the IBM colleagues had no time to deal with the requirements confronting the DEHOMAG.
The result of the construction efforts in Germany was the DEHOMAG BK tabulator, which was announced in March 1933.9 This machine was a break-through for punched card data processing in Germany. The long requested cross-footing feature finally made this tabulator capable of going automatically through the complex wage accounting procedure, to give just one example.
9 Kistermann, The Way, Fig. 7.
Cross-Footing
Let me explain cross-footing. If numbers in different fields of a punched card are to be added (totaled), each of them must be transferred into its own counter (accumulator) during one machine cycle. The next machine cycle then reads the next card and the new data in each field is added only to its assigned counter. This is the same procedure as footing a column of numbers. Conventional tabulators were not able to transfer and add numbers contained in different punch fields on the same card, because reading these fields was a parallel process, as was the addition in the assigned counters. Once the card had been read, there was no way to alter the results.
Controlling the way the card was processed after it had been read, actually required additional cycles, called intermediate cycles (later named program steps), between the reading of two successive cards. This could only be achieved by temporarily stopping the card feed. Since this was caused anyway by automatic group control when a new group identifier was detected, this card-stop was used in the newly constructed machine as the ideal occasion to start intermediate cycles during which data would be transferred from one counter to another. The data was added when transferred. A number of these transfers, that is, cross-footing operations, could be done in parallel by the new machine.
More flexibility in printing was achieved by breaking up the permanent connection between the printing devices and the counters (which had been in existence since 1908). The printing devices could now be wired separately on the plugboard. It was also no longer necessary to punch negative numbers as complements.
Cross-subtraction allowed the automatic conversion of the counter's content into its complement during the transfer to another counter.
The new wiring possibilities resulted in more complex wiring on the built-in plugboard. This would not have been efficient in the long run. Fortunately, a short time after the DEHOMAG BK tabulator was announced, the exchangeable plugboard was introduced. This allowed new applications to be wired apart from the tabulator.
The DEHOMAG BK tabulator was so great a success that DEHOMAG decided to enhance it once more. One year later the BK was finally able to compute the compound interest on bank accounts, previously a very tedious type of work for bank clerks. This machine, called DEHOMAG BKZ, also encouraged the company to continue with this development line, which eventually led to the DEHOMAG D11.10
10 Friedrich W. Kistermann, ''Multiplication, Division, and Printing with Punched Card Machines," Annals of the History of Computing 19:4 (1997) 67–69.
2—
The DEHOMAG D11 Tabulator
The D11 is the culmination of forty years of evolution of the Hollerith punched card system. 11
It began in 1895 with the freight account application in the New York Central and Hudson River Railroad, when Herman Hollerith transformed his 1890 electric tabulating system (pure counting and tabulation) into a business-oriented one, the Hollerith punched card system. This can also be called the Hollerith data processing system to emphasize that there is, in principle, no difference between using electromechanical and electronic devices.
The D11 is an operator-oriented machine with the printing device in front of the operator, the card feed on the left and some counters visible on the right side of the machine. The exchangeable plugboard, which holds the
"program," is also on the right side of the machine. The summary punch stands on its left.
Let me turn to a man, Hans Gross, who played an important part in the construction of the D11 after he had worked with Ulrich Kölm on the BK. Although he was not an inventor like Kölm, he assisted him in his work, and complemented him during the development of the tabulator. Unfortunately, we know very little about Gross short life. He died at the age of 34. Because he was the manager of the circuit design department in 1932, it can be assumed that Gross was instrumental, not only in the construction of the BK, but also in the construction of the D11. At that time, Kölm was deputy manager of production and in charge of manufacturing the BK and later the D11. He set up a factory, which was opened in January 1934. Unfortunately, very few traces could be found to support our conjectures concerning Hans Gross. So they may be seen as "informed speculation."
In principle, the D11 was designed with commercial applications in mind.12 Only one motor drives the machine through the main shaft, which in turn drives the aggregates, such as the card reader, the counters, the print unit, and other parts. The transfer of power occurs with a gear transmission which drives the cam shaft, among other items. This shaft provides the timing that controls the electrical actions that can occur in the various cycles (card feed cycle, intermediate cycles, etc.), provided these actions have been "ordered" by the program on the plugboard.
The card feed is connected to the main shaft via a clutch, which provides the means to disconnect the card reader, if the work of the machine is to be stopped by the group control unit. Then, intermediate cycles, instead of card cycles, are started to execute a program, which is wired on the exchangeable plugboard. These
intermediate cycles are mechanical as well as electrical, because relays are being switched and counters turned, as directed by the plugged program. During machine operation, the exchangeable plugboard is a permanent part of the machine and controls its entire operation. Consequently, this plugboard is comparable to a stored
program.
11 Kistermann, The Way, Fig. 10.
12 Kistermann, The Way, Fig. 8.
All machine cycles (the card cycle, the intermediate cycles etc.) are mechanical motions, that is, the motor makes exactly one rotation during one cycle. Intricate timing mechanisms, such as cams, clutches, gears and other devices, ensure that all operations follow in orderly sequence, that is, a counter will not accept data outside of its allotted time, the printer will move only when results are available, and so on.
With the help of a sorter, the cards are grouped into an order suitable for the report to be generated. The feeding of the card deck is controlled by the group control. A change in the group identifier starts up to nine
intermediate cycles, during which cross-addition and cross-subtraction can be done, even several in parallel. It is also possible to print intermediate results.
The tabulator can start a multiplication with up to eight digits in the multiplicand and six digits in the multiplier, and the result can be used in subsequent operations. Certain operations can be made dependent on a condition, for instance, the punching of a zero-balance can be prevented or a negative total can be printed in red. And a
The tabulator can start a multiplication with up to eight digits in the multiplicand and six digits in the multiplier, and the result can be used in subsequent operations. Certain operations can be made dependent on a condition, for instance, the punching of a zero-balance can be prevented or a negative total can be printed in red. And a