The float-feed type of carburetor which was then coming into prominence in automobile work proved at that stage of its development to be totally unsuitable, as the slight but constant tremor of the aerodrome frame, when the engine was working at high speeds under a heavy load, caused the float to act as a pump and periodically flood the carburetor. This resulted in an irregularity of action of the engine which at times injured not only the transmission shafts, gears, and frame, but the engine itself by the serious pounding which occurred. A form was next tried in which the gasoline was fed in through the valve seat of a lightly loaded valve which raised whenever there was suction in the inlet pipe, the amount of gasoline fed being regulated by a pin valve. Later there were built several shapes and sizes of tanks filled with absorbent material, which was saturated with gasoline and the surplus drawn off before starting the engine. Some of these tanks were provided with a jacket through which [p225] a portion of the exhaust gases was passed in order to compensate for the cooling of the tank caused by the evaporation of the gasoline. As a result of these tests it was found that a type consisting essentially of a tank filled with small lumps of a porous cellular wood (tupelo wood) which was initially saturated with gasoline, and into which the gasoline was fed through a distributing pipe as rapidly as it was taken up by the air, which was sucked through it by the engine, gave the best results. Instead of jacketing this tank, the cooling effect due to evaporation was compensated by drawing the somewhat heated air from around the engine cylinders up through the loosely packed lumps of wood. When tested in the shop this type was found to give such a very uniform mixture that the engine ran as smoothly and regularly as an electric motor, the vibration in no way interfering with it, and even when the sudden change from a state of rest to one of rapid motion through the air was imitated by suddenly turning on the carburetor the blast of several large electric fans from various angles, it was found to have no appreciable effect on the running of the engine, thus indicating that the trouble which was experienced with the model aerodrome in the trials of 1901 was not likely to be repeated with the large aerodrome. Somewhat more than a dozen carburetors of various forms were constructed before this last type was devised, but this proved so satisfactory that there were never thereafter any carburetor troubles. In fact, as will later appear, a carburetor of this type kept the engine on the large aerodrome running at full power not only when the aerodrome was in a vertical position in the air, but also after it had turned completely over on its back. [p226] CHAPTER IX THE QUARTER-SIZE MODEL AERODROME