Calibration of the - 7 r D. M. Poole and 



Emery Settling Tube ' W. S. Butcher 



for Sand Analysis 



i.e., 0.8^. 



SiDlittinf: Error : In running a sample split in l6 different portions (see 

 fig. 5) I the splitting error will be given by extracting the square root of • 

 the error of the settling tube from the total error of the test. In this case 

 the total probable error was 3»9? and the splitting error 3«87o. Figure 4 is a 

 graph of the cumulative freo^uency curves of the l6 samples. In a similar 

 series of saiirples (not shovm) where the splitting procedure shovm in figure 5 

 was not followed, the splitting error was 6,275. 



Because of the large discrepancy in the value of the splitting error in 

 the above tvro samples, the splitting methods were checked again. Eight splits 

 were taken from a sample by combining alternate quarters as sho\>ni in figure 5> 

 and eight splits from the same sample without combining. The splitting error 

 in the first case (combined) v;as l.ofo, and in the second case O.Qfo. The 

 graphs of these two samples are shovm in figures 6a and 6b. Further tests 

 with 32 combined and 30 non-combined samples (not sho^jm) gave splitting errors 

 of Z.kyo and 1.873 respectively. The sand used for these tests had a larger 

 median diameter and was more nearly homogeneous than the sand used in the 

 other tests. It seems obvious from the discrepancy between the series of 

 tests that the splitting error lias not yet been fully investigated. 



The lack of correlation betv;eon splitting method and splitting error is 

 probably due to insufficient data. As vra-s pointed out in the comparison of 

 Ludwick's work and the results of this paper, the splitting error probably de- 

 pends in part on the degree of homogeneity of the sand. The combination sand 

 used in these tests is a non-homogeneous mixture and p. larger splitting error 

 would be expected than in a normal beach sand. Since the probable error is a 

 measure of the variability of a series of tests, we would also expect a greater 



