George Julius presented a paper to the Institution of Engineers Australia on Thursday May 13th 1920 describing these systems, when a machine that had been built and tested capable of supporting 1,000 terminals and a sell rate of 250,000 bets per minute was demonstrated. This image is extracted from that paper and following is some of the accompanying text from it:
The paper describes the development of a machine that has been built in Australia capable of meeting such requirements and of recording records received from as many as 1000 independent operators, and at speeds as high as 4000 a second. What George has described here, is a large scale low response time system. These machines are capable of both printing and issuing tickets, and at the same time recording the issue of such tickets when issued in great numbers and simultaneously by a number of selling machines.
The machine to be installed therefore, must automatically record from instant to instant the total sales on each horse, and the grand total of all sales, and must display these figures in such a way that they may be easily legible to the public. What George has mentioned here regarding the instant recording, makes these real time systems. This last requirement necessitates the use of very large counters or numerators, as the figures require to be legible from a distance of at least 200 feet.
This latter condition necessitates the use of counter wheels of large diameter , even as much as 2 feet, and as the speeds at which they are required to revolve is sometimes great, and the inertia, however lightly they may be constructed, considerable,(sic) they cannot therefore be started or stopped suddenly. Further, also, in such installations it is necessary to locate many of the ticket-selling booths at a considerable distance from the adding machine, which necessitates the use of electric power for the transmission of the records from the selling machines to the recording machine. What George has mentioned here sounds like a distributed network.
Here, again, difficulties arise, as the first requirement of a totalisator is absolute accuracy, and the use of electric transmission obviously introduces a possible weakness which has to be guarded against. Here George introduces a concept that is so important in the digital computer era particularly in real time systems - Reliability Thus, an electric cable may break, insulation may fail, magnet coils may burn out or short or there may even be a complete interruption in the supply of electric power to the machine. A complete system of safety gear has therefore to be introduced, which will only permit of the issue of a ticket at any booth on any horse if the electric connection between that selling machine and that horse is in order, and electric power available.
This may be more briefly expressed by saying that the whole installation must be so arranged that no ticket can possibly be issued without its issue being correctly recorded and vice versa, that no "record" can be transmitted and recorded without the corresponding issue of a ticket.
One more factor also is of importance. The whole equipment has frequently to be worked at very high pressure during selling operations, and the liability of faulty operation of the ticket-selling machines is thus greatly increased. The design of these equipments has, therefore to be such as to make them as nearly " fool proof " as possible.
The foregoing will have made clear the very peculiar and somewhat exacting conditions that have to be met in order to ensure a successful solution of the problem.
The first and most essential factor is the obtaining of a mechanism which will add the records received from a number of independent operators. What George has mentioned here regarding the number of independent operators, makes these multi user systems. This has been done in two ways. The first method, which has met with a certain measure of success in small equipments, depends upon the release of a marble or steel ball whenever a ticket is issued. Further reference to the marble tote has been removed as it is irrelevant. It can be seen in the Mechanical Aids to Calculation chapter of this website. The other system of "collective adding," as it may be called, depends primarily upon the use of a group of super-imposed epicyclic gears. Such a group is shown diagrammatically in Figure 8.
I have only presented this small image of Figure 8 as it does not represent any actual equipment and is only for instruction purposes.
In describing this gear, reference will only be made to the process of " addition," but it is obvious that the gear is equally applicable to subtraction. The gear, as shown, is arranged to receive and add records from six operators, and to show the total upon the total wheel marked " T," by rolling this total wheel towards the left, as shown by the arrow. For convenience let it be assumed that a movement of the wheel "T" of 1/4 inch to the left represents the issue of one ticket. The six selling machines are connected to the wheels "A," "B," "C," and "D," and to the racks "E" and "F" respectively. The double racks "P," "Q," "R," and "S " are not connected to selling machines, and are merely portion of the adding mechanism.
Suppose the wheel "A" to be connected to a 10s. issuing machine, and to be so arranged that the issue of each 10s. ticket causes it to roll 1/4 inch to the left, as shown by the arrow. If the operator of the selling machine connected to this wheel "A" then issues a ticket, the wheel will travel 1/4 inch to the left, rolling upon the momentarily fixed rack "R," and thereby moving the rack "P" 1/2 inch to the left. The teeth on the upper face of the rack "P" will then obviously cause the total wheel "T" to roll along the momentarily fixed rack " Q," and thus to travel, as a whole, 1/4 inch to the left. It is seen, therefore, that the movement of the wheel "A" 1/4 inch to the left will of itself cause the total wheel " T" to move 1/4 inch in the same direction.
It is obvious, however that the gear as shown in diagram No. 8 could not be used, because the racks would have to be of impracticable length, and in practice the racks are replaced by bevel wheel. The various records are then made by the rotation of the wheels A, B, C, D, and hence T about their axes, instead of by the lateral translation of these axes.
This arrangement is shown in Figure 9 (the corresponding parts in the arrangement shown in Figures 7 and 8 being "lettered" the same), and its operation should readily be followed, it being merely necessary to change the motion of translation of the gears shown Figure 8 to one of rotation of the gears shown on Figure 9.
Looking at my comments made regarding some of George's statements above, putting them all together we have a Large Scale, Low Response Time, reliable, Real Time, networked multi user system in 1920! I will also add the observation that at the time I retired in 2012, 92 years later, a transaction rate of 4000 per second was considered good performance for digital computer based totalisator systems.
The part of George's paper that is pertinent to this website is presented in the Mechanical Aids to Calculation Chapter of this website.
Thanks to the Institution of Engineers Australia for allowing reprints of any portion of the Mechanical Aids to Calculation publication.