A MULTIPLE UNIT STEERABLE ANTENNA 395 



along the time axis. (A discussion of other ways of averaging will be 

 given later.) To permit the maximum time interval during which the 

 restoring torque had negligible effect a balancing current was applied. 

 A telechron motor marked time intervals of 15 seconds with a bell. 

 The switchover betw^een "1" and "2" (Fig. 35) was made and the 

 fluxmeter restored to zero reading at the ring of the bell. The signal 

 meter reading (calibrated in decibels) was maintained fairly steady 

 with the automatic gain control, and could be averaged accurately 

 by eye. 



Test Results 



The first measurements on transatlantic signals were made on 16 

 meters in September, 1935. Tests were made on the British Post 

 Office Station GAU (18,620 kilocycles) between noon and 2 p.m., 

 E.D.S.T., on September 16, 17, 18, and 19. At our request the trans- 

 mitter operated on reduced power, presumably fifteen decibels less 

 than normal. The propagation during these tests was characterized 

 by low angles and slow fading. The angular discrimination of the 

 MUSA for these low angles is so slight that it was decided to use only 

 one branch and defer the question of diversity in these first tests. 



Owing to the power reduction and to the fact that the period from 

 September 15 to 18 inclusive was disturbed some of the data had to 

 be obtained without antenna pads. Table II summarizes the results 

 of the measurements. 



When thermal noise is a contributing or predominating factor, the 

 signal-to-noise ratio of the six-antenna branch must be compared with 

 both No. 1 and No. 6 antennas in order that the line loss may be 

 accounted for. \\ ith thermal noise predominating, the difference be- 

 tween the improvements referred to No. 1 and No. 6 should of course 

 equal the line loss of No. 6. It may be shown that the arithmetic 

 means of the two improvement ratios (voltage) should give very 

 closely the improvement corresponding to random static (10 log 6 

 = 7.8 decibels). In Table II the arithmetic means of the improve- 

 ment ratios are, therefore, called the equivalent improvement. 



During these tests the indicated angle of arrival was from one to 

 three degrees (such low angle determinations are not trustworthy 

 within perhaps two degrees) and the receiving branch was set corre- 

 spondingly and not altered during a test. The fading on No. 1 and 

 No. 6 antennas was usually but not always unlike. Adjacent antennas 

 always showed substantially synchronous fading. 



A sample of the plots from which the figures in the table were ob- 

 tained is shown in Fig. 36. This shows the noise readings reduced to 



