334 HELL SYSTEM TECIIXICAL JOURNAL 



was ciedurcd. The method of doing this is set forth in Ajipendix D. 

 This work resuUed in the following table: 



Relative Number 



Signal of Letters for a 



Elements Given Number of 



per Letter Signal Elements 



American Morse (two current values) 8.26 74 



Continental iMorse (two current values) 8.45 73 



Ideal (two current values) 6.14 100 



Continental Morse (three current values) 3.77 163 



Ideal (three current values) 3 . 63 169 



The column in the above table headed "Relative Number of Letters 

 for a Given Number of Signal HIements" makes possible direct com- 

 parison with the results predicted from the formula as gi\en in the 

 table which preceded. It will be noted that the ideal three-current- 

 value code gives an increase in the number of letters for a given 

 number of signal elements as compared with the ideal two-currenl- 

 value code which is in fair agreement with the theoretical ratio of 

 1.58:1. It will also be noted that the Continental three-current- 

 value code which is actually in use in the case of submarine cables 

 appears to come quite close to the ideal. In the case of the Conti- 

 nental and American Morse codes, however, w^here only two current 

 \alucs are used, the results fall short of the ideal, the ratio between 

 the results actually obtained and the ideal being approximately 1.4:1. 

 The reason for this is that a certain proportion of the possible speed 

 is sacrificed in order to make it possible to read the signals by means 

 of a sounder instead of recording them. For instance, the dash has 

 been assumed to be approximately three times as long as the dot. If 

 the signals were mechanically formed at the sending end and recorded 

 at the receiving end, it would be possible to make use of markings 

 1, 2, 3, etc., signal elements long, as well as corresponding spacings. 

 The ideal codes were so constructed. 



It will be seen that the figures deduced for the Continental Mt)rse 

 and the American Morse are substantially identical for two current 

 values. This result probably does not correspond with practise; 

 it is thought that the difference in speed between these two codes is 

 considerably greater, say on the order of 10 or 15 per cent, in favor of 

 the American Morse. The discrepancy is due partl\- to the fact that 

 no account has been taken of figures and punctuation marks in the 

 present comjjutations and partly to the fact that the assumptions as 

 to relative lengths of space is not strictly in accordance with jiractise. 



From the foregoing, it is seen that there is a two-fold gain in chang- 

 ing from the two-current-vahic American or Continental Morse 

 codes to the tliroe-current-\altie Coniinenlal code. In the first 



