CONSTANT FREQUENCY OSCILLATORS 95 



potential of 150 volts and a filament current of one ampere were 

 selected as reference points. The frequency under these conditions 

 was noted. A change to a different operating condition, say 140 volts 

 plate potential and 1 ampere filament current, was made and the fre- 

 quency measured as rapidly as possible, whereupon the operating volt- 

 ages were returned immediately to their reference values and the 

 frequency rechecked. Special care was taken to keep the room tem- 

 perature very constant, but, even so, the heating of the parts of the 

 oscillator circuit by the operating alternating and direct currents was 

 sufificient to affect the frequency to an undesirable extent, requiring 

 that the readings be taken with unusual rapidity in order to return 

 the voltages to normal before the changed operating currents could 

 appreciably affect the temperature of the coils, tube elements, and 

 other parts of the circuit. 



The final results, however, are consistent enough to be representa- 

 tive of what can be accomplished, and the two sets of curves shown in 

 the figures bring out a result which was found to hold throughout the 

 investigation ; namely that the higher the coefiicient of coupling in the 

 coil used to secure feed-back, the less critical was the value of the 

 stabilizing impedance. Thus, in Fig. 20 the coupling was as tight as it 

 was possible to produce by winding the primary and secondary simul- 

 taneously upon a tube to form a single layer solenoid, w^hile the coil 

 used for Fig. 21 was made by removing about half the turns from the 

 secondary of the same coil, thus providing for a step-down in voltage 

 as well as a decrease in coupling. A possible explanation for the less 

 critical adjustment required with tight coupling lies in the fact that 

 the tightly coupled coil satisfies the condition for "unity coupling" as 

 given by (26) over a range of frequencies whereas the loosely coupled 

 coil satisfies the condition at only the frequency critically determined 

 by the stabilizing capacity. 



This would appear to indicate that the type of stabilization de- 

 scribed in the theoretical part of the paper under the unity coupling 

 concept oflers certain practical advantages over those types where the 

 stabilizing element is placed between the tuned circuit and the tube, 

 as in the Colpitts oscillator of Figs. 5, 6, and 7, for example, where no 

 magnetic coupling whatever is employed. Experiments with the Col- 

 pitts circuit have shown, however, that when the capacities in the 

 tuned circuit are made large relative to the inductance, a very satis- 

 factory degree of stabilization may be secured at the lower frequencies 

 where interelectrode capacities may be neglected. A close inspection 

 of the theory of stabilizing the frequency of the Colpitts oscillator 

 shows an argument analogous to that of the coupled coils, namely that 



