46 THE ADDRESSES, LECTURES, ETC., OF 



son's galvanometer. If now the cable, which remains always 

 charged, is put to battery on board ship, an extra charge will 

 flow in at the condenser, and affect the galvanometer at the land 

 end, enabling him to speak to land while the insulation test con- 

 tinues not through instruments, it is true, but through a delicate 

 galvanometer, which may, however, suffice for conveying the 

 necessary instructions. I am not quite satisfied in my own mind 

 whether this method would work with sufficient distinctness to be 

 practicable in paying out a very long cable ; but for a short cable 

 I am quite certain that it would work well. It is one of those 

 methods which, although not yet practically tested, I have thought 

 right to mention. 



TESTING FOR FAULTS. If, in paying out, a fault appears 

 within a short distance of the ship, then we could pull back and 

 remedy the fault, and go on. But, unfortunately, it happens 

 sometimes that a fault appears after the cable is laid. If it 

 appears immediately after the cable is laid, it must be a rupture, 

 or the cable has been very badly manufactured ; but in course of 

 time, when the cable has been exposed to chafing near shore, or to 

 an accident', faults will appear, and we must be prepared with 

 methods to determine the position of such faults, with a view to 

 their repair. This is one of the most important and most beauti- 

 ful branches of electrical testing. There are different conditions 

 and circumstances to be observed. If we have, first of all, a fault 

 in a cable of which both ends are accessible, as would be the case 

 if the cable contained two conductors, and in uniting the two 

 conductors into one enabling us to operate upon both sides 

 from the fault, we would be in the condition represented by 

 diagram Fig. 5, Plate 2. We then get the determination of the 

 position of the fault by finding two equations, the one giving the 

 resistance of the whole conductor I = x + y and the other giving 

 the proportion between x and y, namely a : Z> = x : y, and in de- 

 veloping x and y from these, we have x = , and y = 



v* This method is the best, being the only one in which the 

 a ~\~ o 



resistance of the fault is wholly eliminated. 



Another condition is, in dealing with a single cable, when the 

 fault still permits us to speak through from end to end. In that 



