514 



BELL SYSTEM TECHNICAL JOURNAL 



Thus, the length of the beat periods may be read directly from the 

 photographic records with an accuracy of 0.001 second, which de- 

 termines the length of the 10 second periods with an accuracy of one 

 part in ten thousand. Thus a variation of one division on the photo- 

 graphic record corresponds to a relative variation of one part in ten 

 billion between the two frequencies, compared during an interval 

 only ten seconds long. The whole number of revolutions of the scale 

 may be determined readily by auxiliary means. 



The two graphs in Fig. 24 show typical variations between two 

 crystal oscillators operating under rather unfavorable conditions. 

 One crystal was not in its sealed bell jar and one circuit was only 



2 10 



^ 20 

 5 10 



,>; -20 



o • • • 



• •—(I 



^^t_?_! . ! ,-, — *_^ 



DO 50 100 150 200 250 300 350 4-00 



TIME IN SECONDS 



Fig. 24 — Relative rates of two pairs of crystal oscillators showing small 

 random variations in frequency. 



partially shielded and was exposed to draughts of varying temperature. 

 Even under these conditions, however, the variations from the mean 

 did not exceed one part in 10^ during the test. 



The checking methods just described are intended primarily to 

 indicate more or less rapid changes in frequency. Slow changes, and, 

 of course, the absolute rate, can best be determined in terms of standard 

 time. This is done, as previously indicated, by checking the rate of 

 a clock controlled by one of the crystals against radio time signals. 

 Unfortunately no long checks have been obtained as yet in this way, 

 but several tests made over periods of a few days indicate a constancy 

 of rate in the order of 0.01 second a day. 



The measurements made so far indicate that the frequency of a 

 crystal controlled oscillator such as described when suitably controlled, 

 may be expected to be constant to at least one part in 10^ over periods 

 of seconds or over periods of days. It is hoped that it will be possible 

 in the near future to present accumulated data on the performance of 

 the frequency standard system described. 



