THE CONSTANTS OF THE CUP ANEMOMETER. 
793 
from this that here the coefficient of friction increases with the pressure. ] see no 
mechanical reason for this, but must accept the result, for the difference is clearly 
marked. The outward pressure was resisted by a single roller bearing on one side of 
a plate fixed to the shaft ; this must have tended to make the latter press more on its 
bearings, and as it acted on the plate 1’6 inch from its centre, the friction was needlessly 
great. Therefore on May 18th it was replaced by two opposite ones, larger, carried 
by the shaft, and bearing on the back plate of the frame 0‘5 inch from the centre of 
the shaft. This reduced the friction by 0"6 of its first amount. The measures were 
now repeated. AYith P=56 lb., F = 277'5 grains ; with P=70 lb., F=392, each also 
a mean of three. Flere, however, f must not be subtracted ; for with the two 
rollers the pull raises the shaft sensibly from the outer bearing. These give for the 
F 
first 5 ’2 16, for the second 5 '8 9 5. The simplest mode of representing these is the 
formula F=P These give for the first set (which were used till No. 79) 
£=5-8232; y— 0T272, and for the second £=2-5010; y=0-0509. 
All the constants of the equation (C) except G have been already given ; I give it 
here, and with it the factors for V" for each instrument. 
No, 
I. 
G = 5"4605 
P" = Y 2 xL -1 (8-56260). 
No. 
IT. 
5-0232 
L- 1 (8-46338). 
No. 
III. 
5-0482 
L' 1 (8-53793). 
No. 
IV. 
5-0099 
L' 1 (8-40158). 
No. 
Y. 
5-5965 
L- 1 (8-60018). 
(24.) I suppose this variation of the friction is somehow connected with the rollers, 
but both Coulomb’s experiments and railway experience indicate that rolling friction 
is simply as the pressure. 
(25.) On the whole, I consider that these friction measures, though perhaps not so 
uncertain as those of AY, are yet sufficiently so to increase materially the difficulty of 
drawing accurate conclusions from these experiments. I expected that increasing m 
by additional friction would give a wider range to the coefficients of the equations of 
condition, and this is so. Bat experience shows that the advantage thus gained is 
neutralised by uncertainty when the value of v is small, for then the latter is greatly 
affected by the irregularities of friction to which I have already alluded. This pro- 
bably arises from the momentum of the apparatus not being sufficient to overcome any 
casual increase of friction. But it is also possible that as the friction when a body is 
started from rest is greater than when it is in motion, the passage from one state to 
the other may be gradual through a certain small range of v. 
(26.) Lastly, I shall describe the means which I employed to determine directly the 
coefficient a of equation (I.) and the results which they gave. The outer end of a 
strong clock-spring was attached to an arm of the anemometer set horizontal, the 
inner end to the horizontal arm of the machine : a circle divided to 100 parts was 
5 i 2 
