48 BELL SYSTEM TECHNICAL JOURNAL 



established at + 15 volts and broken in its turn. Possibly because of 

 the continually increasing distance, the arc is not reestablished, and 

 the voltage builds up with oscillations of a frequency characteristic of 

 the line wire insulated at both ends until it reaches — 300 volts a 

 second time and another spark passes. This time only one arc is formed, 

 and the recovery of the voltage starts from the positive side of the 

 zero axis. The current surges corresponding to the voltage collapses 

 of Fig. 11 are shown in Fig. 12. Here the first pulse represents a 

 sparkover which formed only one arc. As the current from the line 

 reached about 4 amperes it was checked and the conducting arc was 

 broken (possibly by being extended laterally into the region of cooler 

 metal). The second pulse shows the current of a sparkover which 

 formed two arcing periods. 



These phenomena are shown in more detail in Figs. 13 and 14 which 

 show the voltage and current of a sparkover forming only one arc, and 

 15 and 16 which show the wave forms when two arcs are formed. 

 Note that the frequency of the current oscillation is that of the line 

 grounded at one end only (the impedance of the load relay being high 

 at this frequency) and is about half that of the voltage oscillation which 

 is that of the line open at both ends. Oscillations of both frequencies 

 may be found in the line at a distance from either end. 



Corresponding observations may be made of the occurrence of 3, 4 

 and 5 arcing periods, the pattern followed being about the same. 

 The higher the voltage at sparkover the more arcing periods; an odd 

 number of arcing periods is followed by a recovery of the voltage from 

 the opposite side of the zero axis from that of the voltage before spark- 

 over, an even number by recovery from the same side of the zero axis. 

 The arcing periods are individually complex, having superposed on 

 them oscillations believed to be due to the relay structure and the 

 leads to the oscillograph which are too fast to be resolved photo- 

 graphically by the means available. These oscillations may be ob- 

 served visually by using higher sweep speeds and reach frequencies of 

 250 megacycles. 



While the arcs ordinarily do not exceed a microsecond in duration, 

 they are probably an important factor in determining contact erosion, 

 as several hundred may occur at each contact opening. 



As may be seen from Fig. 3 the sparkovers continue to occur, the 

 successive voltage breakdowns corresponding to the normal sparking 

 potential as the contact separation increases with time (with some 

 irregularities due to residual ionization in the gap) until the separation 

 is finally so large that the energy remaining in the load relay cannot 

 charge the line to the breakdown voltage. At this stage the line dis- 



