the conductor thus passes over the spark as arc to ground. As soon, however, as the conductor is discharged and at ground potential, the arc between conductor and ground ceases, since there is no voltage left to maintain it, and so the conductor disconnects from ground. The conductor then charges itself again to its normal Y potential and during the inrush of the charge momentarily the potential builds up to double voltage. Thereby a spark again passes between conductor and ground, discharges it again, opens after discharge, again causes a spark to pass, and so on. So a series of successive sparks occur between conductor and ground, discharging the conductor by currents that momentarily rise to very high values, the discharge current of the capacity of the conductor against ground, over a path of practically no resistance. Each spark discharge sends out an impulse or traveling wave, and so a spark discharge between conductor and cable armor, or in the same manner an arcing ground on an overhead transmission line, as for instance caused by a broken insulator, produces a continuous series of impulses or traveling waves, which follow each other with the rapidity of charge and discharge of the cable or the line, that is, many thousands per second, and so give what has been called a recurrent surge. In a long-distance transmission line, the frequency of the recurrent surge is usually somewhat lower than in an underground cable system, but still thousands of impulses per second.

It is interesting to note that no lightning arrester in commercial service to-day protects against such a recurrent surge, but with such a recurrent surge discharging over the lightning arrester it very rapidly destroys itself, usually by conflagration, and where the lightning arrester is not destroyed by the recurrent surge, it is because the discharge voltage is higher than the surge voltage; that is, the arrester does not discharge the surge, but lets it pass into the station; in other words, it does not protect the station, but is inoperative.

The frequency of oscillations occurring in electric circuits varies over an enormous range: from low frequencies, very little above alternator frequency, up to hundreds of millions of cycles per second, and the effect of the oscillations in the system varies accordingly from the relatively harmless static displays, brush discharges, streamers, sparks, and so forth, of extremely high frequencies, down to the disastrous high-power, low-frequency,

short-circuit oscillations, in which, even in 10,000-volt systems, currents of many thousands of amperes may surge, with voltages approaching 100,000, and with which no protective device can cope, which does not have unlimited discharge capacity, that is, contains no resistance whatever in the discharge path.

LIGHTNING PROTECTION OF ELECTRIC CIRCUITS

From the preceding considerations it follows that the problem of protecting electric circuits from lightning is two-fold: I. To guard against high-potential disturbances entering the circuit from the outside or originating in the circuit.

2. To discharge harmlessly to ground whatever high-voltage phenomena may appear in the circuit.

From atmospheric electric disturbances, complete protection can be secured by putting the circuit under ground, or, where this is not feasible, to put the ground over the electric circuit. This means the use of grounded overhead wires. The overhead ground wires thus protect the circuit the more completely, the more they realize a complete shield interposed between line and sky. While complete protection would require a system or network of grounded conductors above, and also below, the transmission line, very good protection in most cases is secured by a single ground wire of good conductivity, installed well above the line; and my opinion is, that in no place of an electric transmission system can money be more efficiently spent than in securing good overhead ground wire protection.

To guard against the appearance of internal lightning requires constant watchfulness in the design, construction and operation of the system, to avoid all conditions that may lead to the formation of oscillating arcs. Thus poor contacts, loose joints, masses of insulated metal near high-potential conductors, and so forth, should be carefully avoided.

The disturbances that have to be taken care of by the lightning arresters proper, are steady accumulation of static pressure; impulses or traveling waves, and oscillations or surges, occurring singly or in groups and of frequencies varying between many millions of cycles, and ordinary machine frequencies; and recurrent surges, that is, impulses and oscillations, usually of high frequency, following each other in very rapid succession, usually thousands per second.

It is necessary that the discharge over the lightning arrester should occur with the least possible disturbance to the system; that is, the discharge current should be as small as permissible without causing a voltage rise due to the resistance of the discharge path. At the same time, the protective devices must be able to discharge practically unlimited currents, that is, currents of the magnitude of the momentary short-circuit current of the system. This obviously requires that the protective devices should have no appreciable resistance in the discharge path. Any lightning arrester containing series resistance obviously fails to protect as soon as the discharge current is so large that the ohmic drop across the resistance becomes serious, and the maximum discharge current, which may occur, is the short-circuit current of the system, that is, extremely large.

Three types of protective devices are at present available. I. The circuit is connected to ground by a single spark gap set for a voltage exceeding the normal operating voltage by a safe margin: the so-called "horn gap," or goat-horn lightning arrester. As soon as the voltage rises beyond the value for which the spark gap is set, it discharges, and the system is short-circuited to ground, until the arc rises and gradually blows itself out. As this requires an appreciable time, motors and converters have usually dropped out of step, and the generators broken synchronism, that is, the system is shut down and has to be started up again. This type of protection, therefore, is not particularly favored in systems that require reasonable continuity of service, but, if used, is rather considered as emergency device in addition to other arresters and then adjusted for much higher discharge voltage. A reduction of the current over the horn gap by series resistance is not permissible, since it correspondingly reduces the protective value, as explained above, and the arrester ceases to protect against a high-power surge. While such surges are relatively infrequent, their destructiveness is such that protection against them is especially needed. Fuses in series with the horn gap, if they open slowly, would still shut down the system, and when opening short-circuit surges over a path of zero resistance, of very high current, the short-circuit current of the system may be disastrous. Obviously, the use of series fuses requires a multiplicity of spark gaps to give continuity of protection.

2. The type of lightning arrester now almost universally used is the multi-gap arrester, which short-circuits the system for one-half wave only. It consists of a large number of spark gaps in series with each other, between metal cylinders. As now designed, different sections of the gaps are shunted with different resistances, for the purpose of affording equal protection against all frequencies, and adjusting automatically the resistance of the discharge path to the volume of the discharge; as for instance, discharge slow accumulations of potential over a very high resistance, short-circuit surges over a path of zero resistance, and so pass a discharge with the minimum shock on the system. The operation of the multi-gap-which, by the way, is suitable only for alternating-current systems-depends on the non-arcing character of certain metals. Metals of low boiling point, as mercury or zinc, can not maintain an alternating-current arc, but the arc goes out when at the end of the half wave, the current falls to zero, and a very much higher voltage is required to again start an arc for the next half wave.* Alloys of such metals, usually zinc, with metals of high melting point, as copper, are used as terminals in the multi-gap arrester.

occurs.

A discharge over the multi-gap arrester short-circuits the system for the rest of the half wave during which the discharge At the end of the half wave, the current falls to zero, and the reverse current can not start; that is, the circuit of the arrester is opened.

A short-circuit on the system for a fraction of a half wave does not interfere with the operation of synchronous apparatus ; that is, the operation of the system is not affected by a discharge over the multi-gap arrester.

In a large system, the short-circuit current is very considerable, its power, and so the heating effect produced by it, enormous. The energy, and so the heat produced by the shortcircuit current during the fraction of the half wave that the discharge over the multi-gap arrester lasts, is moderate, due to its very short duration, and can easily be absorbed and radiated by the arrester; so that even if lightning discharges rapidly follow each other for some. time, they can be taken care of by the arrester with moderate temperature rise: assuming a

*See paper American Institute of Electrical Engineers Transactions, 1906: Light and Illumination.

vicious thunderstorm, in which lightning flashes succeed each other practically continuously, several per second. Each discharge causes a short-circuit over the lightning arrester, varying in duration from nearly a half wave-if the dischargs occurs at the beginning of a half wave-to practically nothing-if the discharge takes place near the end of a half wave-that is, on an average, for one-half of one-half wave, or one-two-hundredand-fortieth second, in a 60-cycle system. Therefore from two to three lightning discharges per second would still short-circuit the system over the multi-gap arrester only for one per cent of the total time, and the heating effect caused by a short-circuit during one per cent of the time can be taken care of by the arrester for a considerable period.

Let us see, however, what happens to the multi-gap lightning arrester in case of the appearance of a recurrent surge, as an arcing ground, that is, discharges following each other in rapid succession, thousands per second. The first discharge, passing over the lightning arrester, short-circuits the system for the rest of the half wave, and at the end of the half wave the arrester functions properly, that is, opens the circuit. At the next moment, however, at the beginning of the next half wave, the next oscillation of the recurrent surge again discharges over the arrester, and so again short-circuits. That is, with a recurrent surge, the multi-gap arrester at the end of every half wave opens the circuit; at the beginning of the next half wave, the next oscillation of the recurrent surge short-circuits again. So far as the effect on the operation of the system and the heating of the arrester is concerned, a recurrent surge causes a permanent short-circuit on the system, except that at the beginning of every half wave, for a short period, the circuit is opened and free for the appearance of disruptive voltages elsewhere, and so apparently, simultaneous with the short-circuit, destructive high potentials may appear in the system. The heating effect of the short-circuit current, which occurs at every half wave, rapidly destroys the arrester. In such cases, to save the arrester, it is customary to insert a series of auxiliary gaps, which are thrown in by the blowing of a fuse shunting them, and raise the discharge voltage of the arrester so that the recurrent surge does not pass over it. It is obvious that in this case the arrester ceases to protect the system against the recurrent surge; but if