946 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



repeater. It should be kept in mind that this assumption would not be 

 valid if the contribution of one element to the deviation pattern of the 

 repeater becomes dominant, say, larger than all the other elements 

 combined. 



In the L3 system, the first stages of equalization are spaced about 25 

 repeaters apart. In accordance with the above considerations, a distribu- 

 tion of individual element variations, the average of which is controlled 

 close to the nominal, will result in cumulative gain deviation which will 

 be substantially smaller than if no restrictions were placed on the aver- 

 age. Thus, a desired objective for critical L3 component elements is to 

 provide a stabilized production process giving a distribution of individual 

 values (a) having an average that is maintained consistently close to a 

 desired nominal, and (b) having a pattern of variation around the 

 average that is Normal (or nearly so). If this is attained, comparatively 

 wide limits for the individual units are acceptable. Furthermore, assem- 

 bly of component elements into amplifiers can be made on a random 

 basis, and the problem of maintaining equipment in the face of replace- 

 ment of parts failing in service will be greatly simplified. 



From the very nature of the over-all problem the best approach to 

 this objective has appeared to be through the application of statistical 

 quality control methods, both in the design of and in the production of 

 the component elements that are important from an equalization point 

 of view. 



2.0 Basic Features op Distribution Requirements 



2.1 general plan 



For each of the important component elements, then, interest centers 

 on closely controlling the collective quality of the product, especially the 

 average of the individual values. This can hardly be accomplished merely 

 by specifying and securing compliance with the usual type of require- 

 ments, expressed as maximum and minimum limits for individual units. 

 Something more is needed. Consideration must be given to ways and 

 means of placing requirements on the distribution of individual values 

 from the successive increments of the product turned out day after day. 



Accordingly, a general plan using quality control methods has been 

 developed, specific features of which will be discussed in this paper with 

 particular emphasis on the intent of certain choices that were made and 

 on the procedures selected to meet the general objective. Further de- 

 velopment work on some of these features may of course be found 

 warranted as experience with them is gained. 



