x machines.

[Illustration: Figure 29.--Section through cabin of the Otis elevator.
Note the pivoted floor-sections. As the car traveled, these floor-sections
were leveled by the operator to compensate for the change of inclination;
however, they were soon removed because they interfered with the loading
and unloading of passengers. (From _La Nature_, May 4, 1889, vol. 17, p.
360.)]

Here then was a design exhibiting strange contrast. It was on the one hand
completely novel, devised expressly for this trying service; yet on the
other hand it was derived from and fundamentally based on a thoroughly
traditional system. If nothing else, it was safe beyond question. In
Eiffel's own words, the Roux lifts "not only were safe, but appeared
safe; a most desirable feature in lifts traveling to such heights and
carrying the general public."[12]

The system's shortcomings could hardly be more evident. Friction resulting
from the more than 320 joints in the flexible pistons, each carrying two
rollers, plus that from the pitch chains must have been immense. The noise
created by such multiplicity of parts can only be imagined. Capacity was
equivalent to that of the Otis system. About 100 people could be carried
in the double-deck cabin, some standing. The speed, however, was only 200
feet per minute, understandably low.

If it had been the initial intention of the designers to operate their
cars to the second platform, they must shortly have become aware of the
impracticability of this plan, caused by an inherent characteristic of the
apparatus. As long as the compressive force acted along the longitudinal
axis of the links, there was no lateral resultant and the only load on the
small rollers was that due to the dead weight of the link itself. However,
if a curve had been introduced in the guide channels to increase the
incline of the upper run, as done by Otis, the force on those links
traversing the bend would have been eccentric--assuming the car to be in
the upper section, above the bend. The difference between the two sections
(based upon the Otis system) was 78 deg.9' minus 54 deg.35', or 23 deg.34', the
tangent of which equals 0.436. Forty-three percent of the unbalanced
weight of the car and load would then have borne upon the, say, 12 sets of
rollers on the curve. The immense frictional load thus added to the entire
system would certainly have made it dismally inefficient, if not actually
unworkable.

In spite of E