502 BELL SYSTEM TECHNICAL JOURNAL 



tance d = Ge\ — Gti will appear as shown. As the repeller voltage is varied 

 from the optimum value the conductance curve will shrink in proportion to 

 cos A0, and the amplitude of oscillation for each value of M will adjust itself 

 to the value corresponding to the intersection of the load line and the con- 

 ductance plot as shown in Fig. 26b. When the load line becomes tangent, 

 as for amplitude F4 , further variation of the repeller voltage in the same 

 direction will cause oscillation to jump from F4 to zero amplitude. Cor- 

 respondingly, on starting oscillation will restart with a jump to Vz . Hence, 

 two sources of conductance varying in this way will produce conditions pre- 

 viously described, which would cause hysteresis as shown in Fig. 26c. 

 The above assumptions lead to hysteresis symmetrically disposed about the 

 optimum repeller voltage. Actually, this is rarely the case, but the ex- 

 planation for this will be deferred. 



Fig. 27 shows repeller characteristics for an early model of a reflex oscil- 

 lator designed at the Bell Telephone Laboratories. The construction of this 

 oscillator was essentially that of the ideaUzed oscillator of Fig. 1 upon which 

 the simple theory is based. However, the repeller characteristics of this 

 oscillator depart drastically from the ideal. It will be observed that a 

 double jump occurs in the amplitude of oscillation. The arrows indicate 

 the direction of variation of the repeller voltage. The variation in the fre- 

 quency of oscillation is shown, and it will be observed that this also is dis- 

 continuous and presents a striking feature in that the rate of change of fre- 

 quency with voltage actually reverses its sign for a portion of the range. A 

 third curve is shown which gives the calculated phase A0 of the admittance 

 arising from drift in the repeller field. This lends very strong support to 

 the hypothesis of the existence of a second source of conductance, since this 

 phase varies by more than 180°, so that for some part of the rangelhe repel- 

 ler conductance must actually oppose oscillation. The zero value phase is 

 arbitrary, since there is no way of determining when the total angle is 

 {n + f)27r. 



Having recognized the circumstances which can lead to hysteresis in the 

 reflex oscillator, the problem resolves itself into locating the second source 

 of conductance and eliminating it. 



A number of possible sources of a second conductance term were in- 

 vestigated in the particular case of the 1349 oscillator, and most were found 

 to be of negligible importance. It was found that at least one important 

 second source of conductance arose from multiple transits of the gap made 

 by electrons returning to the cathcde region. In the case of the 1349 a de- 

 sign of the electron optical system which insured that the electron stream 

 made only one outgoing and one return transit of the gap eliminated the 

 hysteresis in accordance with the hypothesis. 



