950 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



sponsible for systematic deviations which are so undesirable in the L3 

 system. One way to combat or, rather, circumvent large bias is to rely 

 on comparison standards in the calibration of test sets. The task of 

 furnishing and maintaining reliable comparison standards has many 

 pitfalls. It is an art or perhaps a science in itself and is mentioned here 

 only because of its importance to the objective of restricting variations 

 in the process average. 



Considering the effect of both random and systematic errors of meas- 

 urement, it has been found essential to place comparatively tight 

 requirements on the accuracy of the test methods to be used. In many 

 instances meeting these accuracy requirements was made possible by 

 the development of new or radically improved measuring equipment for 

 both laboratory and production purposes. 



2.5 ALLOWABLE MARGIN FOR DRIFT OF PROCESS AVERAGE 



Having determined o-, the basic standard deviation of an acceptable 

 process for a quality characteristic, then the minimum spread of specified 

 limits, to be compatible with the process, would be ± Bo- around the 

 nominal (AT). However, product having this o- could be expected to meet 

 such limits practically all the time provided only that the average of the 

 process were controlled at the nominal. Accordingly, provision was made 

 to allow the process average itself to vary within a band of d= J^o- around 

 the nominal. This allowance is somewhat arbitrary and represents an 

 estimate of relative importance to the L3 system of systematic changes 



NOMINAL 

 N 



•O^A -0.6 A 



MINIMUM 



-0.4A -0.2A 0.2A 0.4A 



DEVIATION FROM NOMINAL 



Fig. 2 — Basis of specified limits. 



0.6A 



0.8A A 



MAXIMUM 



