short circuits of various severity; that is, not only the faulty cable, but its parallel undamaged cable, and the tie cables to other substations, etc., would open, and, while the extent of the shutdown would be somewhat limited, it is still far beyond the extent permitted by reliability of service.

Thus devices become necessary to select a disabled feeder and cut it out without cutting off its parallel feeders or the tie feeders to the substation served by the faulty feeder, regardless of what excess currents these may carry. This is a problem which has not yet been completely solved.

As the result, in general in high-power systems of high standard of reliability the radial system of substation supply is used; that is, each substation is fed by a separate set of cables, and the substations are not interconnected into a network by a system of tie feeders. This radial system, however, is materially less economical in feeder copper than the interconnected network, since the radial system requires for each substation a feeder capacity equal to the maximum power demand of the individual substation, while in the network, by cross-feeding between the substations, the feeder capacity is reduced to that required by the average maximum demand of the substations.

To avoid a shutdown of the substation by a fault in one of its feeders, the different feeders of the same substation are not connected in parallel in the substation, but feed separate translating devices, as transformers and converters, and are paralleled in the substation only on the secondary side of transformer or converter. In case of a faulty feeder, the current feeding back into the fault over the other feeders of the same substation, therefore, has to pass through two sets of translating devices, and this limits. it sufficiently to allow the time limit relay of the circuit breakers to operate and cut out the faulty feeder without opening the other feeders; that is, without shutting down the substation.

However, the economic disadvantage of the radical system remains, and an effective selective feeder relay, which could be relied upon to pick out the faulty feeder and no other, would offer material advantages.

Such a selective device is afforded by the use of pilot cables. Each cable or feeder is duplicated by a smaller low-voltage, threephase cable, which joins the secondaries of current transformers connected into the two ends of the main cable. If the main cable

is undamaged, the same current comes out of it as flows into it, and the connections to the pilot cable are such that in this case the secondary currents would be in opposition; that is, neutralize each other. If, however, the main cable grounds, current flows into it from both sides, the secondary currents in the pilot cable then add, and the current flowing in the pilot cable operates the relay which opens the circuit breaker. This arrangement is very perfect in operation, capable of cutting out the damaged cable, whether feeder cable or tie cable, without interfering with any other cable, but it has the formidable disadvantage of doubling the number of cables required in the system, and, while the pilot cables are small and of low voltage, they occupy room in the underground ducts which carry the electric circuits in American cities. Thus this method of control by pilot cable is, due to its high cost of installation, very little used in this country.

Another method is that of the "split conductor" cable. Every cable conductor is made of two parts, of which the one surrounds the other concentrically, with some insulation between them. Normally there is no potential difference between the inner and the outer half of the conductor, as they are connected with each other at the ends of the cable. If, however, a ground occurs on the cable, this ground can at first reach only the outer half of the conductor, and a potential difference and current appears between the inner and outer half of the conductor and operates the circuit breakers, through a relay connected between the two halves of the conductor, at either end of the cable.

This method also works very satisfactorily, but has the same economic disadvantage, though to a lesser degree than the method of pilot cables, in that the split conductor cable is materially larger and more expensive than the standard cable. It is therefore used to a limited extent only.

The usual method of taking care of the problem, at least in most cases, is by the so-called " reverse power relay," also wrongly called "reverse current relay."

When a cable grounds, the current at its end reverses; that is, flows into the cable (“ feeding back ") instead of coming out of it. However, this reversal of current by itself can do nothing, as it is an alternating current, and as such has no direction of its own, but only a relative direction to other alternating waves, as that of the voltage. Installing then a wattmeter relay at the end

of the cable,—that is, a relay operated by the action of two coils upon each other, the one coil energized by the current, the other by the voltage: if the current reverses, the voltage remaining the same, the pull of the relay reverses, and thereby closes the operating circuit, which opens the circuit breaker which disconnects the cable.

Such reverse power relay operates perfectly so long as there is any voltage for the reverse current to act upon. If, however, a short circuit occurs at or close to the substation, the voltage vanishes, and with it the reverse power relay looses its pull. To guard against this, the installation of reactances is recommended between cables and substations to give a sufficient voltage drop to operate the relay. However, this is an additional complication.

The reverse power relay is not adapted to guard tie feeders between stations, as in these the current reverses in direction with the change of the distribution of load between the substations.

Thus the reverse power relay does not make the operation of interconnected networks of substations possible, but in the radial system of operation, which is generally used, it is the only device which is generally available economically, and is very satisfactory, with the only exception-which must be realized-that it cannot operate where there is no voltage left.

In overhead transmission systems and networks the problems. of selectivity are essentially the same as in underground cable systems, except that in interconnected networks of distributed generating power the high impedance of the lines often gives an automatic partial selectivity, which cannot exist with the low impedance of cable systems.

Interference by lightning, with high-potential transmission lines, has rather decreased with increasing line voltage, and this is very fortunate when considering the enormous extent of these systems and the resulting certainty of lightning effects. In the present high-potential transmission lines voltages have been reached comparable with the voltage of lightning disturbances; possibly not with the voltage of the direct lightning stroke-but direct strokes into lines are rare-but few lightning-induced voltages reach beyond the insulation strength of modern highvoltage lines. In 100,000-volt lines the insulators are tested for one minute at 200,000 to 250,000 volts, and stand momentarily, for the very short time of lightning, over half a million volts.

Thus it is rare that lightning flashes over or punctures the suspension insulators of our very high-voltage transmission systems. A flashover, with the grounded Y system, shuts down the circuit, often without any damage, while with the isolated delta system it may not even shut down the circuit, but is taken care of by the protective device against flashovers, the arcing ground suppressor in the station. Most lightning voltages incapable of destroying the line insulation run along the line until their energy is dissipated or they reach a station, and there they often do serious damage. The most important problem of lightning protection thus has become the rapid damping out of line disturbances caused by lightning, so as to make them harmless before they reach the station. The most effective method heretofore is the overhead ground wire. By its screening effect it lowers the voltage which lightning can induce in the line, but far more important is its powerful damping effect on the line disturbance, the travelling wave caused by lightning, which runs towards the station. As short-circuited secondary to the line wire, the ground wire absorbs and dissipates in its resistance the energy of the travelling wave, and causes it to die out at a rate several times more rapid than is the case in a line which is not protected by ground wire.

Far more destructive than the energy of lightning may be the internal energy of the system. While a lightning stroke may amount to millions of horse-power, at a duration of a millionth of a second, this means only thousands of foot-pounds. In a transmission network of thousands of miles extent a surge of the system may amount to many thousands of horse-power. But even a surge of a hundred horse-power only is liable to be very destructive, as it may be continual, the generator power continually replenishing the surge energy, and a hundred horsepower, during one hour, means 200,000,000 foot-pounds.

Thus the foremost problem is again the protection of the system against destruction by its internal energy, and lightning is dangerous mainly by letting loose the internal energy.

Against damage by breakdown to ground, by overvoltages, effective and complete protection is given by the aluminum cell lightning arrester, so that the problem of overvoltage protection resolves into the economic question, how far the cost of lightning

arresters is warranted by the elimination of the danger of breakdown to ground.

Impulses, high-frequency oscillations, and stationary waves are the most common other dangers.

An impulse is an electrical effect in which voltage and current rise rapidly-with a "steep wave front"-and then gradually taper down and die out. Such impulses are produced by switching operation, by flash over the insulators, by induction from lightning flashes, etc. The danger from impulse voltages lies in the local piling up of voltage due to its steep wave front. Thus, for instance, when a switch is closed and 100,000 volts put onto a line at the moment of closing the switch, 100,000 volts suddenly appear in the line at the switch, while perhaps five feet away the line voltage is still zero. Gradually-with the velocity of light, or 188,000 miles per second-the voltage spreads over the line as an impulse. Suppose now such impulse reaches the terminals of a transformer near the source of the impulse, where its wave front is still very steep. When the full impulse voltage reaches the first transformer turn the second transformer turn is still at zero voltage, and the full voltage of 100,000 comes on the insulation between the first two turns. While the transformer winding is insulated to stand 200,000 volts to ground for one minute, and momentarily still much higher voltage, normally the voltage between adjacent turns may be only ten volts, and with all the extra insulation between the end turns, even if we make this insulation to stand 1000 times the voltage to which it is normally exposed, it would be far below standing 100,000 volts. Thus the danger from impulses is that they produce voltages across small parts of the circuit, single turns or coils, which are often many thousand times the normal voltage existing in this part of the circuit; thus they may be far below the total circuit voltage, and thus would not discharge over the overvoltage protective devices. or "lightning arresters."

Fortunately such steep waves fronts rapidly flatten out in the progress of the wave along the circuit, so that their danger is largely limited to the immediate neighborhood of the origin of the impulse.

Assuming now that we would, by a condenser in shunt to the circuit, bypass the energy of the impulse for only one-millionth of a second. During one-millionth of a second the impulse travels