THE OBSERVATIONS. 



23 



The gate was set by bringing an index mark 

 opposite a graduation mark on a scale, the two 

 marks being on brass plates in contact. The 

 gate was controlled by rack and pinion, and 

 considerable force was necessary to move it. 

 The limit of error may have been 0.002 foot. 

 The ordinary error is believed to have been less 

 than 0.001 foot. An error of 0.001 foot in the 

 sotting would cause an error of 0.002 ft. 3 /sec., or 

 2-J-jr of the medium discharge. 



The determination of head is subject to an 

 accidental error and a systematic error. The 

 accidental error pertains to the adjustment of 

 water level in the high trough, by means of the 

 valve at the pump, with observation of the 

 tube index of water level. It was possible to 

 give this adjustment a refinement comparable 

 with that of the hook gage, but in practice that 

 refinement was not attained, because a close 

 watch was not kept on the index. It was 

 found by experience that the fluctuations of 

 level (occasioned by fluctuations of the electric 

 current supplying power to the pump) were 

 small, and they were usually neglected, a prac- 

 tical calibration of the valve at the pump being 

 arranged so that it could receive the proper 

 setting for each setting of the discharge meas- 

 uring gate. The ordinary error of the adjust- 

 ment of the head is estimated at 0.003 foot, 

 which would occasion an error in the discharge 

 of 1 in 4,000. 



The remaining possibility of error is con- 

 nected with the history of the apparatus. At 

 the time of the calibration of the measuring 

 gate the laboratory occupied temporary quar- 

 ters. In its removal to permanent quarters 

 there was a measurement and readjustment of 

 the vertical distance constituting the head. 

 Also, for the work with the long trough the 

 measuring gate was transferred to a replica 

 of the high trough, which may have differed 

 in some particular affecting the constants. As 

 the work of calibration was at no time re- 

 peated, there was no check on the errors which 

 may have been thus introduced. In a gen- 

 eral way, they are probably of the same order 

 of magnitude as the errors of adjustment 

 of water surface. It is believed that all other 

 errors affecting discharge are small in compari- 

 son with that connected with the measuring 

 gate. 



The vertical width of the aperture by which 

 discharge was regulated was 2 inches. The 



head, measured from the middle of the aper- 

 ture, was 6.0 feet. The horizontal dimensions 

 of the aperture, during experimental work, 

 ranged from 0.1 inch to 6.0 inches, and the cor- 

 responding discharges are given in the follow- 

 ing table : 



TABLE 2. Gate readings and corresponding discharge*. 



THE FEEDING OF SAND. 



The fact that the hourglass has been used to 

 measure time suggests that the flow of dry 

 sand through an aperture may be uniform. 

 Such a flow was not tested in the laboratory 

 because the plan for experimentation required 

 that sand should be used over and over, and 

 it was not practicable to dry it. The hopper 

 was a device intended to produce a uniform 

 flow of wet sand. Moist sand will not flow 

 through a small opening; but if enough water 

 is present to more than fill the voids, adhesion 

 is overcome and flow takes place, as in a quick- 

 sand. The freedom of the flow depends on the 

 amount of water. It was found difficult to 

 maintain a uniform condition in the hopper. 

 Another difficulty arose from clogging of the 

 openings, and this was occasioned by shreds of 

 wood fiber and similar impurities in the sand. 

 The second difficulty was largely obviated, 

 after a time, by making the openings larger 

 and fewer; but the hopper feeding was at best 

 not sufficiently uniform to be used in measur- 

 ing the load carried by the experimental 

 stream. 



For all experiments in which a large quantity 

 of ddbris was carried, the material was fed to 

 the current by hand and was measured in the 

 feeding. A small box of known capacity was 

 filled with the material and emptied into the 

 current at regular intervals timed by a watch 

 or clock. If the interval was long, the meas- 

 ured unit was dumped on a sloping table above 



