58 Transactions. 



are hopeless for the night for one, the other desists from further 

 observations. The accuracy with which these comparisons are 

 made is far beyond the accuracy that is possible in a time- 

 determination : while the probable error of the latter is, say, 

 O'Ol s., that of the former is generally less than 0-002 s. 



The exchange on the cable is similar to that just described 

 of arbitrary signals. The chronograph here is replaced by the 

 paper fillet of the cable service. It is scarcely necessary to 

 observe that nowadays signals (messages) on the cable are not 

 read b}' means of deflections of a small mirror, interpreted on 

 an opal glass scale by means of a reflected beam of light, but are 

 read from the fillet of paper on which a siphon records in ink the 

 deflections. As the current is very weak the siphon is not in 

 direct contact with the paper, but, by an ingenious vibrating 

 de\dce, it deposits a tiny drop of ink at very brief intervals. A 

 cable message looks like a profile of the Rocky Mountains, the 

 ups and downs having an interpretation like the dots and dashes 

 in the Morse system of telegraphy. From experience it is found 

 impracticable to have the clock recording directly on the cable 

 for interpreting signals sent or received. However, it is neces- 

 sary to have a time-measuring scale on the fillet. We accom- 

 plish this by attaching another siphon to the frame of the cable 

 instrument. This one is quite independent of the cable. It is 

 actuated by a long vertical rod attached to the horizontal arm 

 of an ordinary sounder, and connected to the siphon by a silk 

 fibre. This latter siphon drags an ink-line on the fillet. The 

 sounder is put in circuit with the clock, and hence every time the 

 clock or chronometer breaks circuit the sounder makes a sharp 

 break in the line on the fillet, and a time-scale is obtained close 

 to and parallel to the zero-line of the cable-siphon. By pro- 

 jecting vertically these recorded clock-breaks on to the cable- 

 siphon record, we can interpret in time the arrival or departur 3 

 of a signal. We must know, however, the relative position of 

 the two siphons. The signals are sent with one of the two 

 cable-keys (on cables there are always two keys, one for sending 

 positive and the other for sending negative currents). To the 

 lever of the cable is adjusted another lever which is in the clock- 

 circuit. It is so adjusted that the moment the cable-key makes 

 contact — that is, sends a current into the cable — at the same 

 moment the clock-circuit is broken, thereby both siphons record 

 the event simultaneously, and the parallax between the two 

 siphons is obtained. As a check on the value thus obtained for 

 the parallax, a slight tap is given to the frame carrying both 

 siphons, thereby disturbing both, and the parallax obtained. 

 By the above arrangement, when sending signals we have two 

 records on the fillet, one by the clock-siphon, the other by the 



