170 BELL SYSTEM TECHNICAL JOURNAL 



range of variations with respect to the optimum input level to the 

 detector depends on the method by which the sensitivity of the latter 

 is adjusted. Two methods have particular advantages: In one of these 

 the detectors are adjusted for a nominal received level which is made 

 the same for all the channels. This is the method usually employed 

 for cable circuits. It permits adjusting the detectors at any time 

 without reference to the particular line with which they are to be used 

 and without the assistance of an attendant at the distant station. It 

 will be evident that the departures from optimum line gain must then 

 be reckoned from this nominal net circuit-gain, which in the illustra- 

 tion is shown as being .9 db below the actual mean value. This dis- 

 crepancy is principally due to imperfect equalization of the line. In 

 general it is least in the neighborhood of 1000 cycles and increases 

 progressively as one goes away from this frequency. Furthermore, the 

 standard deviation of the distribution curve increases generally in the 

 same manner, so that for a 12-channel system the condition illustrated 

 is perhaps the most unfavorable one. 



A second method of lining up is to adjust the detector sensitivity so 

 as to give unbiased operation with signals transmitted from the distant 

 station and with the line loss whatever it happens to be at the moment. 

 On the average, and in the long run, the effect of this procedure is to 

 restore the symmetry of the variations, but the standard deviation is 

 multiplied by a factor equal to the square root of 2. This is because 

 the occurrence of a given departure from the optimum level is then 

 further conditioned by the particular net gain which happens to exist 

 at the time the detector is adjusted, and the chance of a gain of this 

 particular value is given, of course, by the same distribution which has 

 just been discussed. Most of this increased latitude in variations can 

 be eliminated, however, by seasonal adjustments; a procedure which 

 is evidently of no help when the first method is followed. This second 

 method has been found useful on open-wire circuits because the 

 average net loss of telephone channels over these facilities depends 

 somewhat on their frequency allocation and varies more widely than 

 is the case with cable circuits. 



If no provision were made to compensate for these line-variation 

 effects the result would be a rapid change in bias as the level at the 

 input of the detector departs from its optimum value. This is 

 shown for a typical telegraph channel by the dotted line in Fig. 4. 

 By the use of a level compensator associated with each individual de- 

 tector a great improvement may be obtained, however, the bias varia- 

 tions being reduced to those illustrated typically by the full line in the 

 same drawing. The resulting changes in teletypewriter orientation 



