TM No. 377 



An experiment was run during calibration tests in the Converse tank 

 (see chapter II) to further examine this source of error „ In an acceler- 

 ation test the OMDUM III meter was towed parallel to the u meter axis,, 

 Over a period of about 14 seconds , the meter was accelerated from rest to 

 about 300 cm sec" 1 and then decelerated to rest again. At velocities 

 above 150-130 cm sec°", the recorder (running at the maximum tape speed 

 of 100 mm sec"" 1 " and full pen darkness) displayed a blurred trace, making 

 visual resolution difficult „ An attempt was made to read the tapes. The 

 results of the data record in the range from 200-300 cm sec™- 1 were disas- 

 trouso At about 150 cm sec" 1 , the range of scatter of the points was 

 almost 10 percent of the mean value » Above 200 cm see" 1 the scatter at- 

 tained about 30 percent of the mean magnitude of the velocity. Below 

 150 cm sec"- the data seemed increasingly well=-behaved, with few or no 

 spurious points appearing in velocity ranges below 110=120 cm sec" 1 ,, 



In the estimate of both the variance and the auto»spectra, the over- 

 all contribution of e?.*rors by a small number of spurious points is mini- 

 mized because of the averaging processes involved,, Since the errors are 

 caused by human reading of the data record,, they tend to be random, except 

 that they do occur mostly at the positive and negative peaks of the wave 

 motions. These errors wovdd most likely contribute a relatively small 

 amount of energy to the higher frequency bands of the spectra c This is 

 because the spurioxis rises and falls of the points occur abruptly during 

 the period of sampling (i<>e 8 , generally at C„2-second time intervals) „ 

 The covariance and cross^spectra,, however, are liable to be more sensitive 

 to spurious fluctuations,, in a manner discussed later in this section e 



Data Tap e Reading -=- The resolution of the tape reading device (Tele- 

 reader) is t03 co\mts per inch., or about 15*87 counts per mm„ According 

 to the manufacturer (Telecomputer, Inc s )j> and also in the opinion of 

 personnel reading charts, the random error in reading the horizontal 

 position of the vertical pulses is well within 5 counts or + o ly'6 vmLo 

 (it was indicated by the service representative that, with practice, 

 one could read within 3 counts j but, for purposes of discussion, assume 

 the former value )„ This means that for a, chart speed of 100' mm sec" 1 

 (which gives the greatest pulse spacing) the time of the p-aJ.se interval 

 may be accurately determined to within + 1„53 milliseconds a Since the 

 voltage pulse spacing is a hyperbolic function of time, the error or 

 uncertainty in the velocity estimate grows larger with increased velo- 

 city,, Figure III-6 shows the error in the estimated velocity based on 

 a constant tape reading error of + O.I58 mm (at 100 mm sec"^ tape speed), 

 equivalent to a time error of + 1«58 milliseconds «, Since the general 

 shapes of all wave meter calibration curves are similar, (e„g e , figure III<=5)j 

 this graph (figure III-6) provides a suitable error estimate for all wave 

 meters o 



The slope of the curve in figure II1-6 is monatonieally increasing (in 

 fact, as the square of the current speed) j thus, at the higher velocities 

 (upwards of 150-200 cm sec" 1 ), errors could account for sizable percentages 

 of the total value,. This study is concerned, of course, with the velocity 



66 



