CARRIER TELEGRAPH METHODS 521 



sistance noise, the noise was introduced into the line through a symmetrical 

 three-way pad, as previously mentioned. The marking carrier level at one 

 input of the noise pad was kept at a constant value. The level of the noise 

 current entering the other input of the noise pad was adjustable, and its 

 r.m.s. value was measured with a thermocouple, permitting the computation 

 of the r.m.s. noise-to-carrier ratio at the output of the three-way pad, since 

 the loss through the pad was the same for both signal and noise. This ratio 

 was used for abscissae in Figs. 18 to 21, inclusive, and the ordinates represent 

 distortion for various arrangements. These cur\^es are useful in comparing 

 the relative noise sensitivities of the different arrangements, because the same 

 noise source was used in all tests. The absolute noise sensitivities of the 

 various arrangements were not known accurately because the band width of 

 the noise source was not known exactly. The band width must have been 

 about 3 kc. because the measured signal-to-noise power ratio of the on-oflf 

 arrangement was 15 db better at the receiving filter output than on the line. 

 Fifteen db corresponds to a power ratio of 31.6, which also should be the ratio 

 of the band width of the noise source to the 95-cycle band width of the receiv- 

 ing filter. In comparing the various distortion vs. noise curves, one should 

 note that certain arrangements have different amounts of distortion when 

 noise is absent, which affects the comparison when noise is present. 



Distortion at 60 Words per Minute 



According to Figs. 18, 19, and 20 the arrangements having the same loss 

 characteristic rank in the following order as regards their insensitivity to 

 resistance noise at 60 w.p.m. and 20 per cent distortion: frequency-shift 

 with limiter, closely followed by frequency-shift without limiter, one-source 

 two-band with limiter, two-source without limiter on a par with two-source 

 with limiter, linear on-ofif, single-sideband with — 6 db spacing carrier, and 

 single-sideband with no spacing carrier. All the tests recorded in Figs. 18, 

 19, and 20 were made on arrangements having linear receiving detectors. 

 No noise data were taken by the writers on the level compensated on-off 

 arrangement. Measurements made by other Bell System engineers on a 

 similar level compensated arrangement follow the general shape of cur\^e C 

 of Fig. 18 for the linear on-ofT arrangement, except for an average displace- 

 ment of about one db to the left. It is not known whether the difference in 

 performance of the level compensated on-off arrangement was due to the 

 difference in detector characteristics or to a difference in measuring tech- 

 nique. 



The frequency-shift arrangements tested were all less sensitive to noise 

 than the on-ofif arrangement. When a limiter was not used, this difference 

 was due mainly to differential recombination of the rectified output currents 



