The hull, as at first constructed, consisted of a cylindrical sheathing open in front, through the rear end of which the boiler and aeolipile projected inward, so that the air taken in at the front would be drawn through the boiler and hearth to the exclusion of lateral currents. In the first tests, however, after the hull had been applied, it was impossible to secure a proper rate of combustion, nearly the whole hull being filled with a bluish flame, while only a very small portion of the gases of combustion passed into the coils of the boiler. The remedy for this lay in obtaining an increased draft, and a small stack was, therefore, arranged to carry off the products of combustion. This proved inadequate, and it was only after several weeks of experiment with various types of smoke-stack, and constant alteration of the aeolipile, that it was possible to make the apparatus work [p061] efficiently when it was inside the hull. Finally such a degree of success was attained that the burners could be kept lighted even when the aerodrome was placed in a considerable artificial breeze, created by a blower in the shop. In connection with these tests of the engines and boilers, some method was desired, in addition to the Prony brake tests, by which the thrust of the propellers when driven by the engines at various speeds could be measured accurately and in terms which would be readily available in judging whether the aerodromes were ready to be given an actual trial in free flight. Such a method was found in the use of an apparatus known as the “pendulum,” which was introduced near the end of 1892, but was not generally used until the end of 1893. After this time, however, this test was made a condition prerequisite to taking any of the aerodromes into the field, and proved of the greatest assistance in estimating the probable outcome of the trials. The apparatus used, which is diagrammatically shown in Fig. 10, was extremely simple both in theory and operation. It consisted primarily of a horizontal arm ‹AC› carrying the knife-edge ‹B› by which it is pivoted on each side on supporting beams not shown. Depending from ‹AC› is the light vertical arm ‹DE›, rigidly joined to it and carrying the lower horizontal arm ‹FG›, all of which are braced together so as to maintain the arm ‹DE› constantly perpendicular to ‹AC›. To this arm ‹FG› the model was rigidly attached with its center of gravity in line with the vertical arm ‹DE› and its weight increased by the addition of properly disposed flat weights, in order to make the angle of lift for a given thrust of the propellers smaller and less likely to interfere with the working of the boiler and separator.