Fabula 



show that the drop-off was not a boundary layer effect, and the drop-off was at- 

 tributed to the separated flow which occurs at the Reynolds numbers in this 

 work. The circiilar-cylindrical sensors were also found to produce a false tur- 

 bulence signal in the form of "wake ripple" and "wake noise" starting at sensor 

 Reynolds number of 40 to 50. The drop-off for the thermistor-film sensor was 

 verified by "interface -penetration" tests and was attributed to an extra-thick 

 glass coating. It was found that the wedge sensor gave abnormal calibration 

 curves in Polyox solutions, presumably due to coating of the hot film by gela- 

 tinous material collecting on the leading edge. Thus most of the solution work 

 was done with conical sensors, which displayed only minor calibration- curve 

 shifts due to the additives. The improvement may be attributed to the better 

 shape and the film location. In other work (6), the frequency response of simi- 

 lar conical sensors was found to be essentially flat out to 200 cps for 50 cm/sec 

 which is close to the mean flow speed in this work. Because the upper frequency 

 of the useful spectral measurements in this work was usually about 200 cps, the 

 conical sensors were assumed to have flat response in both water and in the di- 

 lute solutions. The results do not deny this assumption, but when most concen- 

 trated solutions are used in future work, the frequency -response question should 

 be re-examined. 



In order to provide for comparison of the towing-tank results for water with 

 measurements in air by Comte-Bellot and Corrsin (7), a square-bar, square- 

 mesh biplane type of grid was used with the ratio of mesh width to bar thickness 

 taken as 5.3. Thus the nominal grid solidity is 0.34. The grid Reynolds num- 

 ber, R^, = Uq ^/v, where u^ is the mean speed and v is the kinematic viscosity, 

 was about 40,000 in the towing-tank tests and from 17,000 to 135,000 in the wind- 

 tvinnel tests. The mesh width was chosen as 11.8 cm on the basis of several 

 considerations. It had to be sufficiently small compared to the test section di- 

 mensions to encourage lateral homogeneity of the turbulence. On the other hand 

 a large value was desirable so that the available sensors would accurately re- 

 solve the smallest turbulence scale of interest. This scale can be assigned as 

 follows. The characteristic dissipation wavenumber is k^ where i/k^ is the 

 Kolmogoroff microscale, {v^/e)^^*, with e the time rate of decay of turbulence 

 energy per unit mass. It is desirable to resolve eddies of the corresponding 

 scale, and so l^, the longest scale of the sensor hot film, should be less than 

 about the half wavelength corresponding to k^, i.e., ^r/kj. The corresponding 

 frequency seen by the sensor, \'.^\]^/2n, should be well below the sensor cut-off 

 frequency f^. k^ decreases with distance from the grid, and in this work a 

 typical high value (for x/M % lo) is 30 cm"^ Thus the sensor size condition is 



-tg < 0.1 cm , (1) 



and the cut -off -frequency condition is 



f^ » 170 cps . (2) 



These conditions were satisfied in this work with the conical sensors and mar- 

 ginally satisfied with the wedge sensor, A large value of M was also desirable 

 to make the grid Reynolds number as high as possible. 



44 



