The steam, then, enters the cylinder above the piston-rod by means of a valve, but when it has half-filled the cylinder and the stroke is also half completed, the supply of steam is cut off. But the piston-rod does not for that reason come to a standstill: owing to the expansive force of the steam the rest of the stroke is completed when the steam has occupied twice the space it did at the time it was cut off--that is, when the half-stroke had been made. Having, therefore, now completed its work in this cylinder, instead of being allowed to escape, the steam is conducted to a much larger cylinder than the first, for the steam still retains much of its expansive force. In this second cylinder, the same thing occurs again, but when it is admitted to a third, it has already lost much of its pressure. It does its work, and having come through the third cylinder has thus undergone “triple expansion.” Now that it has completed the stroke it passes into the surface condenser already referred to, where it is suddenly chilled and converted into water again, and the vacuum thus formed tends to pull the piston back. In the olden days, as we have seen, the vacuum was made by means of the jet condenser, but now it is done by what is known as the “surface” condenser. It is by means of the latter that the fresh water is able to be used again and again. Otherwise, a steamship could only carry enough fresh water for a few days’ voyage, salt water being not used for the boilers, but merely for circulating through the pipes of the condenser to keep them cool. As the steam comes out from the lower pressure it impinges on the sea-water cooled tubes and so falls to the bottom of the condenser as water. It is then pumped into a tank by means of the air-pump, and from the tank it is pumped back again to the boilers by a feed pump, passing on its way through a filter in order that any oil which may have been gathered in the cylinder may be extracted. It now passes through feed-heaters, where it is heated by exhaust steam from the auxiliary machinery, and so when the condensed water again enters the boiler it is almost at boiling point. The advantage of this triple-expansion was found to give greater speed with less expenditure of horse-power. At first the _City of Paris_ was not as fast as she ought to have been, but after the A-shaped brackets, which supported the two shafts, were removed and, instead, the hull at the stern was, so to speak, bulged out to contain the shafts, her speed was found to be 19 knots with an employment of 2,000 horse-power less than she had needed before this alteration.