This technology history page contains a photograph which is one of several belonging to the photo gallery pages which are part of several pages relating to the invention of the world's first automatic totalizator in 1913 and Automatic Totalisators Limited, the company founded to develop manufacture and export these systems. George Julius was the inventor of the world's first automatic totalisator and the founder of Automatic Totalisators Limited and was instrumental in the design of all of this electromechanical equipment.

One of the many large Adders at Longchamps

Click here to go Back
Click on the image to go back to the Photo Gallery

There is no photographer's stamp on this photograph

There is one adder per runner in a race as well as a grand total adder. This is repeated for each pool that the system supported, in this case GAGNANT and PLACE translated from French, Win and Place. Some of these adders are visible in the Longchamps machine room image below. A full sized version of that image can be seen in the first photograph in the Longchamps section of the photo gallery.

An example of Storage memory before Electronics

This image shows one of the many Longchamps central processing system adders. Longchamps had 273 distributed Ticket Issuing Machines (TIMs). Impulses from the TIMs activated solenoids in the adder which tripped escapement wheels and the rotation of these wheels was summed to determine the investment on a particular runner. The adder in this image adds the investments on runner number 5 as indicated by the number on the adder. As the scanners in the following photo in the photo gallery, multiplexed seven TIMs onto each of the solenoids in the shaft adders, a total of 39 escapements were required, split amongst the individual shaft adders, to support the 273 TIMs. The rotation of each escapement wheel is summed by an epicyclic gear train.

There are seven tubular sections visible across the mid section of the adder, spanning from near the front of the adder to the rear, below the bridge like upper sections. The right hand support pillar of the bridge section, passes the mid section of the nearest tubular section and is perpendicular to it. At this point I need to explain that although this whole device is called an adder it has eleven constituent shaft adders, one for each tubular section of the adder. At the rear end of these tubes are the solenoids, escapement wheels and epicyclic gear trains belonging to the shaft adders. The rear of a couple of the adders are visible in the first image in the Longchamps photo gallery section showing the machine room. A reduced version of this image appears below however the full sized image provides more detail. The two men on the left of the image below are looking at the shaft adders in the second complete adder on the left.

The tubes mentioned above are called Storage Screws. These remember the value of summed transactions and as such are a mechanical form of memory. The summed transactions are recorded as angular displacement resulting from rotation of the fast adding shafts at the rear of the adder. I speculate that the name Storage Screw may be the first use of the word storage to imply artificial memory. This angular displacement is read by the slower to respond, inertia imposed acceleration limited parts of the system, at the front of the adder, as they catch up. An image of a partially assembled Longchamps adder, viewed from the rear, showing the adding shafts and storage screws is at the end of the Longchamps section of the Photo Gallery.

Following is a description of one of the eleven storage screws, nine of which are running from the front to the back of the adder. Only seven are visible from this perspective however another two are hidden by the lower counter wheel display panel on the left hand side of the adder. There are another two storage screws on the upper level of the adder running from left to right, the near one being in plain view, with the adding shaft sections visible on the right hand side. In the front adding shaft section of these two, the front row of escapement wheel solenoids are clearly visible and above these, part of the escapement levers can be seen. The epicyclic gear train and the major part of the escapement wheels are hidden by the dust covers sitting on top of the adding shafts.

Back to the front back oriented storage screws. The adding shaft at the back of each storage screw, winds a screw similar to a grub screw up the threaded inside of the storage screw shaft. The slow to respond near side machinery, detects the movement of the storage screw and starts rotating the storage screw shaft in the opposite direction, an action that works to return the storage screw to its rest position. The angular velocity of the storage screw shaft, is controlled by what George Julius describes as Variable Speed Friction Gear. This Variable Speed Friction Gear allows the storage screw shaft, which has been likened to a nut, to accelerate at a rate that inertia allows and eventually achieves an angular velocity greater than that of the storage screw. This occurs either through its maximum velocity being reached or a decline in the betting rate or both, resulting in the storage screw travelling back towards its rest position. The storage screw position sensing equipment, detects when the screw is nearing its rest position and provides feedback to the Variable Speed Friction Gear, to slow the nut down gradually resulting in a gentle stop. This eliminates impact stress to the inertia limited machinery. On the interesting subject of analogies with electronic systems, this sensing, feedback and velocity control system sounds like a closed loop servo system to me.

For the purpose of describing what seems apparent regarding the storage screw position sensing equipment, I will focus on the storage screw on the right hand side of the adder, although this generally relates to all the storage screws, where there are two storage screws paired together, some of the arrangement is the opposite way around. George Julius described the storage screw in a paper he wrote for a demonstration of a system capable of supporting up to 1000 terminals and a sell rate of 250,000 sales per minute, to the Institution of Engineers Australia in 1920. Incidentally George was a major founder of the Institution of Engineers Australia. In that paper he describes the storage screw, which improved my understanding of this memory device, to the extent of being able to write the previous paragraph. I have likened the control of the storage screw to a closed loop servo system. In this analogy, George's description only alludes to the part of this system which would be called the feedback path. The only reference he makes to it is to identify the part which the feedback loop acts on, by naming it Variable Speed Friction Gear. This is the means of controlling the velocity of what George terms the nut, which is the outer tubular body of the storage screw. How the Variable Speed Friction Gear operates, or how the position of the storage screw travelling along the threaded inside of the nut is sensed, is not described. Regarding this part of the system I ascertain what I can from photographs and then resort to speculation about what is probable. Georges paper, presented to the Institution of Engineers Australia, can be read in the Mechanical Aids to Calculation chapter of this website. To read this, click on the first image in this page, scroll to the bottom of the new page and select the Go to the index option in the Navigation Bar and then select the mentioned chapter in the Firstly section of the index.

Part of the storage screw position sensing equipment, which detects the screw position when it is nearing the rest position, consists of the rods extending from the centre of the ends of the storage screw shafts which act upon an associated lever extending vertically downwards from the right hand side of a pivot point above the associated storage screw pulley. The storage screw pulleys can be seen as nine shiny pulleys, with their circular centres facing the front of the adder, in a horizontal row at a level below the lower counter wheels and display panel on the left. The left two storage pulleys are larger than the rest of the pulleys. The rod in our example is seen protruding out from the centre of the right hand storage screw to a position slightly above and to the right of centre of the right hand storage screw pulley.

The storage screw position sensing equipment pivot points seem to have two levers attached to them that work together. In our example, the right hand lever attached to the pivot point above the storage screw position sensing rod, is the longer and its middle is connected to the storage screw rod which moves the lever. The bottom of this lever pushes on an adjustment screw contained in another lever pivoted from below the associated storage screw drive pulley. The upper lever pushes the lower lever and a projection on the lower lever seems to engage a projection on the storage screw drive pulley which seems to have the effect of locking it from rotating when the storage screw has reached its rest position.

The left hand lever has a rod attached to it that disappears rearward down the left side of the storage screw and extrapolating its direction it looks like it connects to a lever that rises to a pivot rod running the width of the adder just below the shelf that also runs the width of the adder behind the lower counter wheel. This rising lever then changes direction 90 degrees at the fulcrum, then travels rearwards and disappears from view behind a cable loom running up the nearest right hand pillar of the adder. This puts it in a position to connect with the activating arm of the right hand Mercury Pot Switch described later.

This switch probably cuts off a circuit associated with the storage screw when the screw has returned to its rest position. The method of achieving this angular velocity control seems to have more than one form when looking at different adders. In this adder it seems the variable friction is achieved by altering the tension on the drive belts. There are tensioner pulleys controlling the tension of the main drive belts. These belts run from the drive pulleys on the main drive shaft, which can be seen across the bottom front of the adder, to the Storage Screw Nut Drive Pulleys arranged at 90 degrees to and located above the Main drive Pulleys. The nine tensioner pulleys can be seen in a row between the Main Drive Pulleys and the Storage Screw Nut Drive Pulleys, near the Main Drive Pulleys.

The main drive pulleys are driven by the large double pulley near the middle of the Main Drive Shaft. I presume this is a double pulley to implement redundancy to eliminate the possibility of a single main pulley belt failure causing a complete adder failure. The twin belts are driven by a twin pulley on a drive shaft underneath the floor. Is this where the computer room false floors originated?

The tensioner pulleys are attached to levers dangling down from their fulcrums and some can be seen extending out to tension their respective drive belts. It is possible that the Variable Speed Friction Gear, referred to by George Julius, in the case of this adder is implemented by these tensioners, however I cannot see from this photograph, whether the extension of these pulleys is controlled by the storage screw position sensing equipment or whether they are just manually adjusted to a fixed position to take up slack originating from different length drive belts or wear.

Another observation about the belt tensioner mechanism. Attached to the adder frame, next to the top of each tensioner pulley, a short distance above the fulcrum, are what looks like threaded rods with springs on them, with a nut holding each spring on its rod. The rods project out in the same direction as the storage screw position sensing rods. It appears that the compression of these springs, determines the force applied to the drive belts, as each of the extended tensioner levers has its corresponding spring rod thread poking out through its nut, implying these nuts are compressing their respective springs to a greater extent than the ones where the threaded rod cannot be seen protruding from its nut. If this is how the tensioner arms are controlled it remains unknown whether these springs are tensioned manually or by the variable friction gear mechanism George refers to.

Neville Mitchell, the best historian of this company I know, has described the mercury pot switch mentioned above, in the Video clips of a working Julius tote chapter under the Technology to shake to heading. The switching circuits he refers to can be seen in this adder. About half way up the adder, there is a shelf that runs horizontally across the adder. On the left hand side, the shelf is hidden by the lower counter wheels and display panel. On this shelf the mercury pots Neville mentions can be seen in pairs with the fork like spans inserted in the pots.

Neville is writing about Giant Drum Julius Tote displays he saw in Thailand however the mercury pot switches he describes are clearly visible in this adder. Neville wrote they had switching circuits for the decade counting from one digit to the next which were mercury pots, with probably about a 25mm diameter pot 30mm deep that was filled with mercury lying side by side, and then a fork like a span, which was the switch, would dip into it, and that would create the circuit to the next decade. And you can imagine how absolutely accurate that was, there was no chance of a dirty contact or a miss-up in a norm transfer.

There are nine of these mercury pot switches on the shelf previously mentioned, which is on the adder in the image. The left hand two are hidden from view. There is one of these for each Storage Screw underneath this shelf. There is a pivot rod visible, running underneath and along the length of this shelf as previously described above. There are levers on this rod, with a right angled bend, which pivot at the bend. In their present position they are oriented with one arm of the lever vertically down and the other arm horizontally rearward. The rearward section of the levers are impossible to see, as they are hidden by the pivot bar, but it is possible to just make out the rearward section of the first right angled lever, on the right hand side of the adder.

The bottom of the vertical part of this lever is controlled by the storage screw position sensing mechanism as mentioned in the text above. The visible rearward end of the horizontal part of one of these levers connects with a small vertical rod rising to join the back of the fork arm, the prongs of which are sitting in the mercury pots. This fork arm is pivoted and this vertical rod acts to raise the fork arm out of the mercury breaking the contact. This is part of the servo system's means of bringing the associated storage screw to its rest position. One of these Mercury Pot Switches can be seen on this adder associated with the upper assembly which has two more storage screws mounted across the adder. The Mercury Pot Switch is beneath the upper storage screws and to the left of the post with the number five and light on it. This is a good contrast with the other Mercury Pot Switches, as this one has the fork lifted out of the mercury and is in the off position.

As Neville mentions, these mercury pot devices can be used as switches. It has dawned on me that in the case of these adders they could have a greater involvement in what I termed the closed loop servo system, associated with the storage screws. Apart from switching, they could also be used as a variable resistance dependent on how deep into the mercury pool the contacts are lowered, reducing resistance between the electrodes the deeper the electrodes are immersed and vice versa. This could vary current in a circuit, as a means of variable control, which is an essential part of a closed loop servo system. This is given further credence by the fact that the immersion of the electrodes in the pots is controlled by the storage screw position sensing mechanism providing the feedback path of the closed loop servo system. If this electrical control could be linked to the screw that tensions the springs that seem to be linked to the tension applied by the tensioner pulleys, then we will have closed the loop of the closed loop servo system, and would have a complete picture of the storage screw's workings.

The Longchamp Machine RoomImage of the longchamps machine room

I have provided the above image here to give an idea of what the title of this page refers to in mentioning this is one of many adders. This is an image of the Longchamps Machine Room which houses the Julius Tote central processing system. Have a look at the light poles on top of the adders and follow them into the distance of the image to get an idea of the size of this system. These were behemoth machines. On the second adder in the right hand row, the light pole runner number on the top of this adder, is legible and it bears the number 29. This means this adder is totalling the investments for runner number 29 which also means that this is the 29th adder. There are as many adders as runners in the race for the Win pool and this is doubled to support the Place pool. There is an additional two adders to calculate the grand totals for Win and Place pools. There will be more adders in the machine room behind the photographer.

When looking at any Julius adder, like this one, it is interesting to remember that you are looking at a piece of equipment capable of parallel processing, which is a modern computing concept. This old system through the Julius Adders, could do something that the digital computers I introduced on the Brisbane racetracks, that replaced these Julius systems could not, parallel processing. I was often reminded of this by the staff I inherited who used to work on the Julius totes. This was achieved through the following feature of design.

If two or more of the Scanners (Time division multiplexers) on the Julius tote, selected TIMs (Ticket Issuing Machines) that had a transaction pending on the same runner, they would activate all the solenoids connected to these scanners at the same time, in the adder associated with that runner. These multiple escapement wheel activations are instantaneously registered by the epicyclic gears in the adding shafts. In the extreme, if every TIM scanned from the respective groups of TIMs on a scanner has a transaction pending for runner 5 for example, which is the runner the adder in this image represents as indicated by the 5 displayed on top of it, and they all happened to be scanned at the same instant, then all of the 39 escapements, mentioned above, would be activated at the same instant and the total investment would be instantaneously recorded by virtue of the epicyclic gear trains.

My new computer totes, although loosely coupled multiprocessors, were not capable of parallel processing. Each processor sequentially processed every bet and this transaction processing was first done by the master computer, which then passed transactions to the slave, to be recorded there, albeit that this happened so quickly that it all looked like it was happening at once. The Julius tote adders could record multiple bets instantaneously, which their replacement, digital computer based totalisator systems could not!