76 KENNELLY-SANBORN— INFLUENCE OF [April 24, 



wires across the test-wire, introduced an element of weakness. After 

 exposure of some days, the test-wire was apt to break at a soldered 

 point. In subsequent out-door trials, therefore, the tap-wires were 

 simply twisted on to the test-wire. These tap-points involve a certain 

 small error due to their cooling effect on the test-wire. It is, there- 

 fore, desirable to make the tap-wire as thin as is practicable, in view 

 of both mechanical strength and electric resistance. 



In a wind, the readings of the voltmeter anemometer are always 

 fluctuating; but a mean value at any moment can always be esti- 

 mated. A heavily damped voltmeter may conveniently be used. 



The rising parabolic curve in Fig. 14 indicates, to the right-hand 

 scale, the watts convected from 15 cm. of the same test-wire at dif- 

 ferent wind-velocities up to 1,100 cm./sec. or 40 km./hr. with con- 

 stant resistance maintained in the wire; i. e., by the method of Fig. 

 12. It is evident that for high wind-velocities, the indirect method 

 of constant test-wire temperature is more sensitive than the direct- 

 reading method of constant current. For lower velocities, however, 

 the direct-reading method is much the more convenient, but requires 

 to be corrected for the wind-temperature. 



In the use of the constant-current direct-reading method, tungsten 

 lamps have an advantage for the regulating resistance of Fig. 13 in 

 that they tend to compensate for changes in the resistance of the test- 

 wire at different wind-velocities ; or to act as ballast resistance for 

 the closer maintenance of constant current. 



Conclusions. 



1. The forced convection of heat from a thin platinum wire at 

 constant temperature moved transversely through the air, varies not 

 only approximately as the square root of the speed, but also approxi- 

 mately as the square root of the air-pressure (or air-density), in 

 accordance with Boussinesq's formula. 



2. At air-pressures below half an atmosphere, the above square- 

 root relation was sensibly departed from. At pressures above 3 

 atmospheres, the relation was also departed from. Although the 

 departures in both cases exceed the limits of observation errors, it is 

 not certain whether they may have been due to imperfections in the 



