1880.] 



J. F. Tennant — On Standard Wcif/lits. 



47 



Section IV. — System of Weif/Jiments. 



I have adopted a uniform sy.stem of weigbment for comparing the 

 weights. Some years ago I made a considerable number of experiments 

 on the species of errors which occurred in practice, and the present system is 

 the outcome : there have been minute deviations, but in all material points tlie 

 procedure has been uniformly followed, and I think it has been successful 

 in eliminating all progressive errors. The principal of these is the ten- 

 dency of the arms of the balance to expand unequally with temperature, 

 but there are otliers which have occasionally been found. I annex specimens 

 of the form I have used iu work. 



The weights to be compared being placed in the pans, a preponderance 

 is given to one side of the balance ; so as to make the resting point, when 

 tlie wliole is in equilibrium, lie on one side of the centre point ; yet so slightly, 

 that the weight used to get the value of the scale, shall deflect the resting 

 point to the other side. In the first example with Oertling No. 1, it will 

 be seen, that with EI in the left pan and Oj in the right, the Right Eider 

 was placed at 1'2 of the beam scale ; in this state the index had its 

 resting point at 7'54 divisions (10 being the middle). Then the weight 

 T-Qi was added to the left side and the resting j^oint became 15 '81 Div. 

 Each resting point is deduced from 4 readings, two low and l^, and 

 two high k-^ and //g. The beam having been carefully released, the 

 first excursion outvvards, and the return towards the scale centre, are neglec- 

 ted ; and the next four readings of the extremes of oscillation taken. The 

 first reading will thus usually be low, if the resting point below ; and high, 

 if that be high : but, when signs of irregularity occur, this may not be the 

 case, as I have always, in such cases, freely omitted readings till the oscillations 



have become regular. Then, supposing a low reading first, ^"^ and 



— ~ would be readings of the resting points, and the sums in the nu- 

 merators have been rapidly formed separately during the work, added, and 

 divided by 8. This has been afterwards checked by ^ ^'^ ^ ' 



of course, when h comes first, the h's take the place of the I's in these 

 formulic, and vice versa. 



We thus have two "2}artial weighnents'^ 



EI =2= Oj + 1-2 at 7-54 divisions and 

 EI + P.oi ^ Oi + 1-2 ^ at 15-81 divisions 



