function of x, Thus, by allowing measurements to be made farther 

 away from the injector, increasing M has the effect of making R, 

 the effective radius of the sampler smaller relative to the useful 

 values of cr „ In the present set of experiments (R s o 03, M = 0.75 

 in.) the effective value of R/M » o 040 seems to be still too 

 large to allow great confidence in the values of cr after the very 

 substantial corrections were made Since, however, the corrections 

 necessary to obtain cr± were less than 10 percent when c^ was as 

 large as o 04, one can perhaps conclude that the scale of the 

 experiment should be so arranged that €T 1 /R>1 5 over a sufficient 

 range in which cr(x) is linear to compute a slope d<r(x)/dx fairly 

 accurately. If R is taken as o 0175 in c , corresponding to the 

 sampler actually used but with velocity matching the stream velocity 

 one must have Ci>0.026in. With a slope d®/dx s 04 ? correspond- 

 ing to L/M s 22, the first sampling station would then be at 

 x s 0,6 in If one wishes to have at least another inch in the 

 linear range, a total of 1.6 in , the mesh of the grid should be 

 about double that in the present experiment, where the linear 

 range was about 3/4 in. longo 



In the present experiments no effort was made to control 

 the rate of discharge of dye from the injector or the rate of 

 removal through the sampler except to try to keep them constant. 

 The rate of discharge should be adjusted to be equal to the free- 

 stream velocity and the rate of removal to be less than or equal 

 to it. The effect upon the flow of having a source and sink at 

 the injector and sampler, respectively, should again be small as 

 long as the diameters of the tubes are small compared with M 



The corrections applied to obtain the true values of a 

 from the experimental curves were based upon the assumption that 

 the true concentration curves were Gaussian curves. This has 

 not been tested. However, a test could be made by comparing the 

 experimental curves with theoretical curves obtained from a 

 Gaussian curve by local averaging as described in the appendix, 

 (i.e. by comparing the experimental g(r) curves with the theoretical 

 g(r) curves). 



- 8 



