THE L3 SYSTEM — EQUALIZATION AND REGULATION 843 



characteristics, temperature affects cable and repeaters semi-independ- 

 ently. In a 4,000-mile system one week is expected to engender as much 

 as 8 db gain change by change in repeater temperature. Cable changes 

 can exceed 100 db per week. Thus these effects must be corrected to a 

 very high order of precision to maintain good television or telephone 

 service. 



These gain changes are not the entire story; when the gain changes 

 in the band it also changes outside the band and usually by an even 

 larger amount. Thus equalization of the in-band gain leaves outband 

 (above 8.5 mc) gain changes which produce in-band delay changes 

 of several microseconds. Because of the difficulty and complication of 

 providing automatic delay equalization it is necessary to equalize these 

 delay changes on a gain basis, by at least partial correction for outband 

 gain changes. Further, since satisfactory pilot transmission is possible 

 only mthin the band, these out-band changes must be predicted from 

 the in-band gain changes. This effect, in itself, indicates the use of 

 equalizers whose individual shapes are those produced by specific sys- 

 tem causes producing a correlated change in many elements. Thus 

 building the regulating networks to match the effects of the individual 

 system causes and matching to 10 or 12 mc rather than just to 8.5 mc 

 permits simultaneous gain and delay equalization. Such a set of shapes 

 is also more accurate because it is matched to the special ways in which 

 the specific system can change rapidly. 

 I In theory the regulating networks could be built to match linear 

 combinations of the cause shapes but there are two difficulties. First, 

 not all of the cause shapes are known accurately or often even roughly 

 at the introduction of a new system into the field. The mixtures of shapes 

 cannot be determined mthout the missing ingredients. Second, a system 

 is not a static design. Experience suggests improvements and thus 

 occasions will arise where one will want to change the correction for a 

 particular cause. If the networks represent mixtures of the causes this 

 necessitates changing aU the networks. If specific networks match specific 

 causes only the appropriate network needs replacement. 



The Computer 



The use of cause shapes leads to a problem to which the computer 

 provides the solution. These cause shapes are broad effects covering the 

 entire band and more. Thus no one pilot is a measure of a specific cause. 

 However by a process equivalent to the solution of simultaneous equa- 

 tions the pilots determine the amounts of an equal number of cause 

 shapes that will restore the pilots to normal. Thus the computer trans- 



