VOLUME EFFICIENCY OF REPEATERED CIRCUITS 531 



maintenance, it is possible to reduce this effect to about ±.03 db per 

 regulating repeater. Other regulation inaccuracies, including imper- 

 fections in the design and manufacture of regulating networks and 

 departure of individual pairs from average characteristics, may intro- 

 duce an additional error of about ±.l db per regulator, this effect being 

 more or less random, however. 



In addition to the residual effects of temperature changes there are 

 variations in the net losses of the circuits due to repeater battery 

 changes and humidity changes. The repeater batteries are usually 

 held to fairly narrow limits and vacuum tubes are tested regularly 

 for emission. The expected change in repeater gain due to an "A" 

 battery change of ±.5 volt is about ±.2 db and for a "B" battery 

 change of ± 5.0 volts is about ±.25 db. 



In office cabling and in the switchboard multiple at the terminals 

 of the circuit there may be a considerable amount of variation due to 

 changes in the humidity. This has been largely taken care of by 

 improvements in the type of cable used (cellulose acetate) and by 

 keeping the lengths of office cable as short as possible. However, a 

 residual variation of about ±.5 db may be expected, a considerable 

 part of w^hich is due to switchboard multiple. 



If the number of repeaters in a circuit is "w" and the number of 

 regulators is "r," the total variations are considered to be about 



Vi = ± V (.5 + .03r)2 + (.25)V + (.1)V -f {.lYn + {.ISyn. 



These items are allowances respectively for humidity variations, 

 regulator lag, finite regulator steps, other regulator errors, "A" battery 

 changes and "B" battery changes. Rearranging the equation, 



I'l = ± V .25 + .1025r + .OOOP/-^ -1- .1025;/. 



In addition to this variation, the probability that the average net 

 loss of a given circuit is not exactly as specified must be considered, 

 so the variation from the specified value is considered to be about 



Vl Vi or 



W = ± V .5 + .205r + .001 Sr^ + .205«. 



Assuming that each of the individual variations from the various 

 sources has an equal probability throughout its range, the probability 

 that the overall variation 1% will be exceeded is about .085, and the 

 probability that the average variation in the two directions of trans- 

 mission (which is of considerable interest in singing or echo computa- 

 tions) will exceed this is still smaller. 



