S2 ROYAL SOCIETY OF CANADA 



The value of g is taken to three places of decimals, allowing a pos- 

 sible error of 1 in 32,000 or -00003. This en-or being under the radical 

 sign, would be halved, or -00002 say. 



It is ditiicult to say what the average eri-or in the head would be. 

 The cord indicator shows a variation of 01 inch, and when noticed this is 

 immediately corrected. The average error would probably be less than 

 half this amount, but taking it at this figure would give an eiror of 

 •0004 at 1 foot head and 00002 at 20 foot head. These would be halved, 

 being under the radical sign, giving -0002 and -00001, respectively. 



The errors in estimating the various factors may, therefore, be 



summed up as follows : 



At highest head. At lowest head. 



Q -00020 -00020 



A -00010 -00010 



t -00005 -00001 



q -00002 -00002 



h -00001 -00020 



Totals -0004 -0005 



The practice throughout was to take two measurements under each 

 condition. ' If the two results corresponded within about 0003, the mean 

 value was taken as correct. If the results differed by more than that 

 amount, a further experiment was made, until two nearly similar values 

 were obtained. 



In the mean values there should not be an error greater than -0002, 

 and from the consistency of the succeeding values for different heads it 

 may reasonably be concluded that that degree of accuracy is attained. 



The orifices used were, in order to give a fair comparison, all of 

 practically the same area, which is that of a circle of half an inch dia- 

 meter or "19635 square inches. 



The results obtained are embodied in the following table : 



TABLE I.— VALUES OF COEFFICIENTS OF DISCHARGE. 



