perhaps not larger than a needle will just pass through, and if the insulation resistance be measured with a battery, it will probably be found to be high. Such high resistance faults are exceedingly difficult to find, and it is therefore usual to put in heavy fuses and burn the fault into as low resistance a one as possible.

Returning to the testing of faults, it is usually an earth that is tested for, a short or open circuit being by its very nature usually self-revealed; the former by a local disturbance, the latter by cessation of supply.

It is not possible to enter here into details of the various methods employed to localise an earth. Speaking generally, a single main of uniform section, without branches, presents little or no difficulty, whether the fault be an earth or a short, and whether the main be single or concentric.

One of the most useful tests for such a main is the loop test. This requires a second conductor, which, however, is usually available. To Το make the test, a considerable current, usually derived from one or two storage cells, is passed through the faulty cable and through another cable running along the same route, the far ends of the two cables being connected together. Across the ends of the same battery is connected a slide wire, to the moving contact of which is attached one end of the high resistance coil of a galvanometer, the other being connected to earth. There is thus formed a Wheatstone bridge, of which the two portions of the slide wire form one pair of arms, the faulty cable up to the fault and the remainder of that cable, plus the second cable, forming the other pair; the galvanometer circuit being completed through the earth and the fault. When balance is obtained, the position of the fault can easily be calculated, if both mains be of the same sectional area, or the cross-section and length of each be known. Fig. 95 will make the connections clear.

The chief advantages of this method are, that it is independent of the

actual resistance of the main, and that it is unaffected by the resistance of the fault, a high resistance merely reducing the sensibility of the galvanometer. It is capable of giving very close approximations to the position of the defect, the Author having known faults localised to within a few yards in several miles.

The accuracy of the test necessarily depends, among other things, upon the degree of fineness of adjustment of the slider, which obviously varies with the length of the slide wire. For a long main, a number of wires are necessary; a very convenient form, devised by the Author for the London Electric Supply Corporation, and since copied by Messrs Elliott Bros., consists of a single wire passed round a series of pulleys, so as to form ten

parallel sections; a number of knife-edges clamped on to the wires serve to exactly determine the length of wire in use. The slider has two motions at right angles to one another, so that it may pass from one section to the next. The arrangement is shown diagrammatically in fig. 96. The object of using a single wire passed round pulleys is to secure uniform tension, and therefore equal resistance, in all the ten wires. Until this device was resorted to, it was found very difficult to obtain equality.

There are numerous other ways of testing based upon the loop test into which it is not necessary to enter here. They differ merely in the way of carrying it out.

Another method frequently used consists in passing an alternating or intermittent current through the fault, and carrying a large coil, in the circuit of which is included a telephone, along the route over the main, the sound ceasing or diminishing when the fault is passed. In some cases, this

gives accurate results, but it is often misleading, owing to leakage currents passing down such conductors as gas and water pipes.

Sometimes a steady current of considerable magnitude is passed through the fault, as in the last method, and a compass needle substituted for the telephone and coil.

Other methods depend on the use of a differential galvanometer, the instrument showing the difference in the current carried by the two conductors constituting the feeder, which difference is equal to the leakage current. This test, which is only applicable to continuous current systems, can be made without interrupting the supply.

The case considered is only met with in feeders. In a distributing main, the problem is complicated by the service lines, the loop test obviously merely showing that the fault is a certain number of microhms away, and not discriminating between the thick, low resistance, distributor, and the thin, comparatively high resistance service.

In the case of a large single network, the difficulties are so enormously magnified that in the Author's experience any attempt at testing, in the usual acceptation of the word, is out of the question. A little consideration will show why this should be so. In the first place, any method depending on the measurement or comparison of the resistance of the conductors cannot give any determinate result, because there are several or many paths to the fault. Again, methods based upon the presence of an excess current on one side of the system must fail, because this excess current may travel through many different paths. Methods depending on an interrupted, alternating, or unusual periodicity current, either alone, or superposed on the ordinary current, are liable to failure for the same reason, and also because of the many opportunities the leakage current has of returning to the station.

The Author's experience has shown him that, for faults of the kind referred to, there is nothing so rapid or certain as the unscientific and brutal method of trial and error comprised in dividing the network bit by bit until the faulty length is localised.

Suppose an earth develops on a large single network supplied from a number of low-tension feeders, how is it to be found? First of all, entirely disconnect each feeder in turn, momentarily connecting a different pole of the network from that on which the fault is to earth, through a fuse and ammeter, and noting the deflection, or, if the earth be a bad one, seeing whether the fuse goes. If the fault be on a feeder, it will be at once ascertained on which it is by the earth disappearing when the particular feeder is disconnected, and it can then be localised by the loop or other convenient test. If the defect be found to be on the network, this is then divided into two approximately equal parts, supplied from two separate sets of plant. The necessity for the apparatus described in Chapter XXI. for disconnecting mains with the current flowing now becomes evident. The

fault will then be on one or other half. The faulty half is again divided up into two portions, one being transferred to the sound half and the other tested for earth as before. The good portion is left connected to the sound half and the process repeated, sections being transferred to the sound half piecemeal until a single section containing the fault is left. This can usually be reduced to a few hundred yards, and the exact position of the fault be localised by any of the methods of testing proper already described. It is of vital importance, when testing in this manner, to be perfectly systematic. If guesses be made, and parts suspected to be faulty be cut off, the result will only be waste of time and delay.

An approximate guide in a large network may sometimes be obtained in towns in which an earth return is used by the Telephone Co. by observing on which lines the indicators are dropped, and the localities they come from, and it has even been thought worth while, in some cases, to put down special lines for this purpose.

A short circuit is, of course, a kind of fault that tends to discover itself. In a high pressure system, in which it is not the practice to resort to much interconnection, the faulty section usually cuts itself out. In a low pressure system in which fuses can rarely be used, it is the practice to endeavour to burn out the short, and in most cases this is successfully done. This drastic course usually enables the fault to be found, its situation being discovered by some unusual occurrence at the seat of the mischief, the most usual indication being smoke issuing from a junction box. Some clue is often given by observing the feeder ammeters, those on feeders going to the affected neighbourhood having the highest readings.

As already stated, this method of fault-finding is brutal in the extreme, and exceedingly unsatisfactory from a scientific point of view; it is nevertheless effective, and is resorted to by those actually engaged in the practical operation of central stations, though there is not an engineer who would not gladly welcome some scientific method that would relieve him of the necessity for such a nerve-taxing series of operations.

CHAPTER XXIV.

RECORDING THE LAYING OF MAINS.

We have now to consider how the record of the mains laid is kept. The Model Provisional Order requires that a map of the area of supply shall be made, and the line and depth of all mains and services, together with junction boxes and other underground works, be marked thereon. Were there no such stipulation, it would be imperative that such a record should be kept, though it need hardly be said that it cannot be kept on a single map.

The matter is really a complicated one, and the system to be described is the outcome of a good deal of care and consideration on the part of the Author, having been brought by him to its present form by a process of evolution extending over a number of years. It is found to meet every practical requirement in a very complete manner.

There are a number of matters to be recorded in connection with a main that is laid. In the first place, its precise position, as determined by its depth and distance from the buildings, must be known at every point, together with similar particulars for every joint box or service connection. Secondly, the size and nature of the main must be known and particulars as to its insulation resistance when laid. Again, the relation in which a given main stands to the rest of the network, and the points at which disconnection can be made, are matters of vital importance. Finally, the distribution of consumers, particulars as to the demand and distribution of the load, including the important question of balancing, are all within the province of mains records.

In the first place, it is convenient, though not essential, to have a single map of the whole area of supply. The best scale is the 2 feet to the mile Ordnance (more accurately 1 inch to 208 feet), as this shows the houses and enables the premises of any consumer to be at once identified. The separate maps are mounted on linen and varnished. If the area is a small one, the sheet map may be mounted on ordinary rollers and hung up; but, if large, it will become a kind of panorama, and may be advantageously mounted on a spring roller fixed at the top, the bottom being also fixed to a roller, provided with a handle and pawl and ratchet, and fixed in bearings near the floor. This map should be kept for reference only; it should not be attempted to lay down anything whatever on it.