In operation, water was admitted to the two cylinders from a tank on the third platform. The resultant hydraulic head was sufficient to force out the rams and raise the upper car. As the rams and car rose, the rising water level in the cylinders caused a progressive reduction of the available head. This negative effect was further heightened by the fact that, as the rams moved upward, less and less of their length was buoyed by the water within the cylinders, increasing their effective weight. These two factors were, however, exactly compensated for by the lengthening of the cables on the other side of the pulleys as the lower car descended. Perfect balance of the system's dead load for any position of the cabins was, therefore, a quality inherent in its design. However, there were two extreme conditions of live loading which required consideration: the lower car full and the upper empty, or vice versa. To permit the upper car to descend under the first condition, the plungers were made sufficiently heavy, by the addition of cast iron at their lower ends, to overbalance the weight of a capacity load in the lower car. The second condition demanded simply that the system be powerful enough to lift the unbalanced weight of the plungers plus the weight of passengers in the upper car. As in the other systems, safety was a matter of prime importance. In this case, the element of risk lay in the possibility of the suspended car falling. The upper car, resting on the rams, was virtually free of such danger. Here again the influence of Backmann was felt--a brake of his design was applied (fig. 38). It was, true to form, a throwback, similar safety devices having proven unsuccessful much earlier. Attached to the lower car were two helically threaded vertical rollers, working within the hollow guides. Corresponding helical ribs in the guides rotated the rollers as the car moved. If the car speed exceeded a set limit, the increased resistance offered by the apparatus drove the rollers up into friction cups, slowing or stopping the car. [Illustration: Figure 35.--Detail of links in the Roux system. (From Gustave Eiffel, _La Tour de Trois Cents Mètres_, Paris, 1900, p. 156.)] [Illustration: Figure 36.--Section of guide trunks in the Roux system. (From Gustave Eiffel, _La Tour de Trois Cents Mètres_, Paris, 1900, p. 156.)] The device was considered ineffectual by Edoux and Eiffel, who were aware that the ultimate safety of the system resulted from the use of supporting cables far heavier than necessary. There were four such cables, with a total sectional area of 15.5 square inches. The total maximum load to which the cables might be subjected was about 47,000 pounds, producing a stress of about 3,000 pounds per square inch compared to a breaking stress of 140,000 pounds per square inch--a safety factor of 46![16]