1 70 The Rev. T. R. Robinson's Description of an improved Anemometei'. 



beincr 73 grains. This was finally replaced by the present mounting in May, 

 1S49. It is shown in Plate V., figs. 2 and 3, where C is the axle, 0-82 inches 

 diameter, D the box of bronze (8 copper to 1 tin); B, five balls of the same, 

 ri2 diameter; I a disc of iron truly turned on the axle ; H an aperture for in- 

 troducing occasionally a few drops of oil, which I find necessary for the lateral 

 action of the balls. They bear both the lateral pressure and the weight ; and, 

 therefore, require only a slight lateral support below, which is given by the 

 arbor of the endless screw. This arrangement shews no trace of wear after 

 more than a year's work, and the total friction is but 53 grains :* the coeffi- 

 cient of that part of it which belongs to the balls, I find to be „^^ of the load. As 

 in high winds there is added to this a lateral pressure, in respect to which the 

 balls do not act quite so advantageously, we may take it i. From this value 

 of the friction, the correction of V is easily computed ;f but it is in some 



* Of this I find that 20-36 belong to the mill, and 32-64 to the registering apparatus: with 

 the new shaft the total friction will be 48-61. 



f For this it is also necessary to know the constant a - a'. I appi'oximated to this as follows : 

 a spring-balance is attached to a cord wound on the axle C, which, as w =o, measures with_/owr 

 cups the force V^ {a -a'). Its slide moves a pencil parallel to the axis of a cylinder covered with 

 paper, and made to revolve by clock-work, on which it traces the curve of time and force. The 

 small anemometer already described gives V by comparison with the time. This V is reduced to 

 that of the large instrument by comparative trials at the time of experiment. It must, however, be 

 remembered, that it is a mean velocity, and that, therefore, the value of a - a' thus obtained, is too 

 small if the fluctuations be considerable. As Fis affected by friction, the first values of a - a' are 

 used to correct it, and thus a more exact result is given by a second computation. By six diagrams 

 I find. 



The second and fourth were marked as doubtful from excessive fluctuations; but as the mean of 

 them and the fifth differs little from that of the other three which were considered satisfactory at 

 the time of experiment, I retain them, and take a - a' = 31-997, or 32 in round numbers. The 



equation (3) becomes, with the values previously given for — and m. 



