522 BELL SYSTEM TECHNICAL JOURNAL 



of the two branches of the discriminator". When a limiter was used and the 

 noise was small compared to the carrier current, the limiter theoretically 

 should have reduced the noise-to-carrier ratio at least 3 db. Therefore, one 

 might expect curves A and B of Fig. 20 to be separated horizontally at least 

 3 db in the region of low noise and to come together as the noise approaches 

 zero. However, since curves A and B are fairly flat in the region of low noise, 

 small errors in distortion measurement may have caused appreciable errors 

 in the horizontal separation between the curves. When noise is not small 

 compared to signal current, it is expected from theory that a limiter would 

 give only a small reduction in distortion caused by noise on a frequency-shift 

 arrangement. This appears to be verified experimentally by curves A and B 

 of Fig. 20. 



The main purpose of testing the wide-band frequency-shift arrangements 

 was to compare their sensitivity to noise with that of the normal-band 

 frequency-shift arrangement with limiter. According to Fig. 20, curves B 

 and D, at 60 w.p.m. the tolerance to noise interference in the wide-band 

 arrangement with low-pass filters having a cut-off frequency of about 503 

 cycles was 2.7 db greater than that in the normal-band arrangement at 10 

 per cent distortion and 1.3 db greater at 20 per cent distortion. This im- 

 provement was unexpected because the wider band admitted more noise. 

 The improvement must have been due to the smaller distortion in the wide- 

 band arrangement when no noise was present. The theoretical difference 

 in noise tolerance between these two frequency-shift arrangements, which 

 had a two-to-one ratio of frequency band width and of frequency swing, 

 would have been 6 db at low noise levels if low-pass filters had been used at 

 the detector outputs of the wide-band channel to cut off components in the 

 detected current which are higher in frequency than those transmitted by the 

 normal-band arrangement^*. However, the low-pass filters used in the tests 

 had a cut-off well above this value in order to permit signaling at very high 

 speeds. Consequently, there was some unnecessary noise passed by these 

 filters which accounts for an increase in noise tolerance of less than 6 db at 

 60 w.p.m. 



In order to verify the fact that low cut-off low-pass filters improve the 

 noise tolerance" of the wide-band frequency-shift arrangement, the 503- 

 cycle cut-off filters were replaced by filters having a cut-off at about 58 cycles. 

 According to Fig. 20, curves B and C, at 60 w.p.m. the tolerance to noise 

 interference in the wide-band arrangements was then 6.1 db greater than in 

 the normal-band arrangement at 10 per cent distortion, and 2 db greater at 

 20 per cent distortion. 



"J. R. Carson and T. C. Fry: "Variable Frequency Electric Circuit Theory- with Ap- 

 plication to the Thcor>- of Frequency Modulation", Bell Sys. Tech. Jour., Vol. XVI, No. 

 4, October 1937, pp. 513-540. 



